# 1. The basics of double/debiased machine learning¶

In the following we provide a brief summary of and motivation to the double machine learning (DML) framework and show how the corresponding methods provided by the DoubleML package can be applied. For details we refer to Chernozhukov et al. (2018).

## 1.1. Data generating process¶

We consider the following partially linear model

\begin{align}\begin{aligned}y_i = \theta d_i + g_0(x_i) + \zeta_i, & &\zeta_i \sim \mathcal{N}(0,1),\\d_i = m_0(x_i) + v_i, & &v_i \sim \mathcal{N}(0,1),\end{aligned}\end{align}

with covariates $$x_i \sim \mathcal{N}(0, \Sigma)$$, where $$\Sigma$$ is a matrix with entries $$\Sigma_{kj} = 0.7^{|j-k|}$$. We are interested in performing valid inference on the causal parameter $$\theta$$. The true parameter $$\theta$$ is set to $$0.5$$ in our simulation experiment.

The nuisance functions are given by

\begin{align}\begin{aligned}m_0(x_i) &= x_{i,1} + \frac{1}{4} \frac{\exp(x_{i,3})}{1+\exp(x_{i,3})},\\g_0(X) &= \frac{\exp(x_{i,1})}{1+\exp(x_{i,1})} + \frac{1}{4} x_{i,3}.\end{aligned}\end{align}

Note

In [1]: import numpy as np

In [2]: from doubleml.datasets import make_plr_CCDDHNR2018

In [3]: np.random.seed(1234)

In [4]: n_rep = 1000

In [5]: n_obs = 200

In [6]: n_vars = 150

In [7]: alpha = 0.5

In [8]: data = list()

In [9]: for i_rep in range(n_rep):
...:     (x, y, d) = make_plr_CCDDHNR2018(alpha=alpha, n_obs=n_obs, dim_x=n_vars, return_type='array')
...:     data.append((x, y, d))
...:

library(DoubleML)
set.seed(1234)
n_rep = 1000
n_obs = 200
n_vars = 150
alpha = 0.5

data = list()
for (i_rep in seq_len(n_rep)) {
data[[i_rep]] = make_plr_CCDDHNR2018(alpha=alpha, n_obs=n_obs, dim_x=n_vars,
return_type="data.frame")
}


## 1.2. OLS estimation¶

For comparison we run a simple OLS regression of $$Y$$ on $$D$$ and $$X$$. As we will see in the following, due to the considered high-dimensional setting the variance of the unregularized OLS estimates is higher in comparison to the double machine learning estimates and therefore the estimates are also less efficient.

In [10]: from sklearn.linear_model import LinearRegression

In [11]: import matplotlib.pyplot as plt

In [12]: import seaborn as sns

In [13]: colors = sns.color_palette()

In [14]: def est_ols(y, X):
....:     ols = LinearRegression(fit_intercept=True)
....:     results = ols.fit(X, y)
....:     theta = results.coef_[0]
....:     return theta
....:

# to speed up the illustration we hard-code the simulation results
In [15]: theta_ols = np.array([0.59318776, 0.45971453, 0.46241367, 0.48078614, 0.47250707, 0.48543412, 0.70513768, 0.68020653, 0.51909541, 0.47951409, 0.37821017, 0.38242628, 0.55801763, 0.49086914, 0.84131015, 0.7710256 , 0.56847739, 0.45701804, 0.52761673, 0.6346921 , 0.54605122, 0.5723868 , 0.54334723, 0.65323525, 0.61125249, 0.43241426, 0.43104578, 0.45377385, 0.50867609, 0.14461668, 0.40294401, 0.43876645, 0.71483579, 0.4399299 , 0.72038435, 0.52232852, 0.35518506, 0.43642775, 0.31654814, 0.55992389, 0.63356066, 0.23340524, 0.67178528, 0.45811377, 0.64384906, 0.44535273, 0.35326843, 0.46432657, 0.72450818, 0.60858665, 0.64181596, 0.5121659 , 0.48217484, 0.40228978, 0.49558718, 0.62924855, 0.51721149, 0.50069418, 0.80139192, 0.34085407, 0.75088008, 0.30519389, 0.66133905, 0.39275145, 0.68758212, 0.55900036, 0.45024447, 0.30445786, 0.4313834 , 0.64066921, 0.47282037, 0.6848716 , 0.21635294, 0.65622022, 0.72828305, 0.63041858, 0.56954612, 0.57376026, 0.66102439, 0.41659409, 0.66243637, 0.4356498 , 0.29548206, 0.46731586, 0.65360673, 0.22210238, 0.6132152 , 0.78867833, 0.48772794, 0.22668944, 0.5905258 , 0.4000885 , 0.51811107, 0.35222288, 0.48031581, 0.58389996, 0.84767486, 0.50579516, 0.47000761, 0.51530275, 0.39533203, 0.62167001, 0.52558447, 0.62164319, 0.36240803, 0.67980513, 0.61039662, 0.30917031, 0.48828101, 0.65861219, 0.57240869, 0.46475721, 0.59948429, 0.644878  , 0.43811496, 0.50672598, 0.37570644, 0.30098601, 0.66809267, 0.35229073, 0.25343975, 0.21620671, 0.62218367, 0.59439647, 0.34133268, 0.47683222, 0.86136912, 0.53148453, 0.35280422, 0.60376521, 0.36941261, 0.59029651, 0.71232415, 0.35911796, 0.71539705, 0.41632781, 0.41994643, 0.27237688, 0.19610283, 0.33579009, 0.45026933, 0.49008255, 0.49895826, 0.53450753, 0.35522258, 0.29714548, 0.47817931, 0.43378291, 0.66762971, 0.69275525, 0.60789843, 0.67337471, 0.37707381, 0.54609838, 0.78122203, 0.7067751 , 0.31857833, 0.70237046, 0.39886057, 0.38740589, 0.55649901, 0.39888709, 0.49761043, 0.57782971, 0.65502679, 0.23982947, 0.39230768, 0.45254724, 0.75551596, 0.61865677, 0.43315527, 0.70034407, 0.48921717, 0.34448065, 0.29063194, 0.62574567, 0.08560904, 0.52672954, 0.51897996, 0.48633659, 0.40545461, 0.61618659, 0.43888745, 0.70117113, 0.56834975, 0.4353438 , 0.48782113, 0.5458835 , 0.69214867, 0.56980536, 0.66733377, 0.32372753, 0.63668482, 0.44151454, 0.47232849, 0.47463123, 0.67928778, 0.52309043, 0.45893484, 0.67134199, 0.36807602, 0.54232177, 0.48355105, 0.43718358, 0.43891791, 0.59764225, 0.09557678, 0.53424475, 0.55336975, 0.41810503, 0.74685723, 0.58437186, 0.44617578, 0.47894358, 0.37700085, 0.38144555, 0.66718062, 0.82580379, 0.58207645, 0.51268775, 0.43436392, 0.76063177, 0.35217197, 0.64642546, 0.46344298, 0.82406445, 0.22547059, 0.44871022, 0.40777041, 0.45978791, 0.58673064, 0.75452511, 0.27714111, 0.16615785, 0.23653097, 0.51675712, 0.37369916, 0.71688425, 0.39651969, 0.53974942, 0.50124734, 0.87057507, 0.57109674, 0.4961726 , 0.16852742, 0.91126084, 0.60095392, 0.89644763, 0.45482427, 0.57563471, 0.40739706, 0.59947253, 0.26911198, 0.6946975 , 0.33859101, 0.46868632, 0.44633092, 0.45329767, 0.48137464, 0.37221182, 0.36714006, 0.59867284, 0.64480537, 0.56642007, 0.60510489, 0.54518583, 0.56786103, 0.53438459, 0.42251166, 0.62091963, 0.71727092, 0.57126156, 0.48766866, 0.42415296, 0.67484828, 0.26253921, 0.47470388, 0.30665878, 0.26964429, 0.44458861, 0.50890811, 0.51881239, 0.83399069, 0.65199939, 0.3599203 , 0.30065112, 0.59042903, 0.60127214, 0.5442253 , 0.38712326, 0.36662413, 0.59696361, 0.53815764, 0.47350178, 0.2172039 , 0.54504514, 0.46664198, 0.3784215 , 0.22411117, 0.46698582, 0.45802382, 0.80279368, 0.72458228, 0.5332414 , 0.66575705, 0.39375685, 0.62272984, 0.60260018, 0.64966364, 0.32517752, 0.2818564 , 0.57051723, 0.28561171, 0.68267698, 0.76572089, 0.46197899, 0.26354168, 0.62811391, 0.48339201, 0.41040701, 0.34407777, 0.74397161, 0.58163655, 0.58848438, 0.4712122 , 0.63299441, 0.47226971, 0.84312757, 0.58690103, 0.55780786, 0.42650571, 0.72423899, 0.42848945, 0.56600457, 0.26434488, 0.32172571, 0.49642333, 0.46473514, 0.49518052, 0.44338712, 0.35435808, 0.64136388, 0.55406214, 0.64207924, 0.31352479, 0.36003016, 0.49612422, 0.50293705, 0.58085204, 0.44952171, 0.65043484, 0.41191937, 0.66542255, 0.44917533, 0.55543017, 0.45743673, 0.47363382, 0.44719089, 0.62406627, 0.4432179 , 0.37999959, 0.67229461, 0.57084708, 0.8213345 , 0.48569308, 0.66836467, 0.45638755, 0.58352417, 0.47701639, 0.53716006, 0.54852805, 0.32644371, 0.28744608, 0.63268009, 0.53282089, 0.59195658, 0.50357241, 0.52336052, 0.35179723, 0.46349958, 0.73028539, 0.42760158, 0.39364331, 0.26469956, 0.49361765, 0.6433476 , 0.58419397, 0.34466566, 0.44078549, 0.47100607, 0.34141449, 0.67467571, 0.52182799, 0.47971572, 0.45100469, 0.4058031 , 0.37551038, 0.71436864, 0.40943507, 0.41077428, 0.42161418, 0.40876852, 0.4326489 , 0.54801259, 0.58330917, 0.50158302, 0.47047853, 0.52006491, 0.47960384, 0.50235396, 0.21686093, 0.57150066, 0.48165407, 0.67596478, 0.65007464, 0.56010447, 0.47598996, 0.67206994, 0.46907106, 0.65062314, 0.34869573, 0.29144174, 0.28864541, 0.72606039, 0.65388605, 0.44821016, 0.38426961, 0.71613355, 0.40370851, 0.37693571, 0.5962973 , 0.53145349, 0.61313976, 0.75058939, 0.52339309, 0.67136156, 0.62170365, 0.53100461, 0.35853528, 0.52784428, 0.41637952, 0.57350769, 0.47613934, 0.60231196, 0.45251648, 0.48900661, 0.42829154, 0.31043151, 0.73706439, 0.71967152, 0.23245767, 0.12313164, 0.38764358, 0.45058064, 0.29000066, 0.44032338, 0.48756876, 0.25742971, 0.55950368, 0.45302682, 0.32273711, 0.70712568, 0.48139585, 0.48239776, 0.60221386, 0.39319688, 0.59877079, 0.56715362, 0.36249042, 0.17848256, 0.16838167, 0.24272911, 0.68725371, 0.49278375, 0.66689222, 0.4923022 , 0.83595761, 0.65645787, 0.36815945, 0.67653854, 0.75860092, 0.32958781, 0.8739171 , 0.65305523, 0.28007832, 0.5835609 , 0.7885053 , 0.43134965, 0.4665327 , 0.29630251, 0.58487731, 0.31402388, 0.37634538, 0.31017821, 0.54411106, 0.58228149, 0.67857093, 0.4662651 , 0.71991177, 0.54267782, 0.54814876, 0.49230113, 0.44344911, 0.54044095, 0.49486078, 0.34239824, 0.82912331, 0.4931028 , 0.6455143 , 0.72408513, 0.53179327, 0.57669499, 0.27941019, 0.70358373, 0.387985  , 0.47858681, 0.57686246, 0.35358592, 0.54423571, 0.52984391, 0.59134525, 0.38404534, 0.34715857, 0.57577872, 0.49203394, 0.36934865, 0.45879214, 0.52323179, 0.50688769, 0.77795903, 0.39970098, 0.8174074 , 0.70994741, 0.42739496, 0.71587723, 0.210805  , 0.36977447, 0.40940471, 0.45365071, 0.63924987, 0.76144695, 0.49626135, 0.42514929, 0.51333049, 0.52736324, 0.78553965, 0.99836658, 0.58560877, 0.53926485, 0.72507555, 0.5727157 , 0.5537641 , 0.53497333, 0.48095033, 0.49375619, 0.60832012, 0.59776148, 0.61076608, 0.4023326 , 0.4892973 , 0.70641856, 0.59018401, 0.58142925, 0.56410325, 0.17480185, 0.34274071, 0.69928325, 0.57447153, 0.57427467, 0.55804701, 0.4394097 , 0.26998491, 0.16007376, 0.59867694, 0.22770188, 0.6246163 , 0.37485096, 0.77129836, 0.55674016, 0.40985446, 0.51728389, 0.55585407, 0.2972799 , 0.36461585, 0.43635738, 0.50688704, 0.57609329, 0.54863157, 0.52743596, 0.43659008, 0.3630138 , 0.58241221, 0.30694678, 0.77483326, 0.38347196, 0.6486276 , 0.65569848, 0.55886832, 0.38642255, 0.21496472, 0.30529768, 0.66269774, 0.43163026, 0.37097861, 0.37842578, 0.46289863, 0.39130125, 0.35586093, 0.47998112, 0.60116294, 0.47902059, 0.54092779, 0.70337639, 0.48894294, 0.38295417, 0.45362429, 0.60171822, 0.48411367, 0.40540698, 0.55843263, 0.60064329, 0.32481347, 0.48805479, 0.37936013, 0.28188033, 0.18012661, 0.52021594, 0.59807101, 0.21777747, 0.63941109, 0.43250054, 0.31089103, 0.63056542, 0.31113406, 0.68914649, 0.38257375, 0.63430732, 0.57387416, 0.42626419, 0.51559958, 0.51423557, 0.47715882, 0.49123188, 0.33479591, 0.47485337, 0.61250023, 0.55982791, 0.4603608 , 0.57060444, 0.53265265, 0.33005468, 0.33534821, 0.50701757, 0.34863287, 0.60156818, 0.36318835, 0.48653985, 0.63495298, 0.31882775, 0.29070672, 0.61081819, 0.60035029, 0.54191366, 0.58069915, 0.86376021, 0.49147657, 0.55636647, 0.74660267, 0.46704973, 0.50933909, 0.38140262, 0.76941082, 0.5923492 , 0.27393903, 0.43008993, 0.67758134, 0.45128826, 0.77510749, 0.35359248, 0.56363219, 0.42443904, 0.68747136, 0.51261231, 0.29115268, 0.45835881, 0.65477146, 0.45888715, 0.63900626, 0.54637549, 0.62018464, 0.55044645, 0.49422625, 0.35617975, 0.51328578, 0.32400516, 0.39488301, 0.20808939, 0.50059824, 0.59516783, 0.53032753, 0.64233057, 0.17814375, 0.71128197, 0.41190639, 0.22443664, 0.42248144, 0.54223629, 0.54977878, 0.49372526, 0.59574847, 0.57766511, 0.40074797, 0.54447921, 0.42402972, 0.37966303, 0.46526625, 0.37060073, 0.42849493, 0.44825793, 0.23121664, 0.34685528, 0.62648812, 0.51122737, 0.45453868, 0.54934909, 0.31938025, 0.37601716, 0.4033514 , 0.48352587, 0.41168559, 0.41846181, 0.55574752, 0.56982557, 0.46600939, 0.40592558, 0.31251982, 0.38358162, 0.29796192, 0.58859424, 0.25974764, 0.6773811 , 0.47792482, 0.37972602, 0.40794382, 0.32190437, 0.56358727, 0.63143422, 0.51043225, 0.79278151, 0.60150745, 0.61067032, 0.48205424, 0.35165069, 0.63014378, 0.81859016, 0.3736024 , 0.58413088, 0.27355234, 0.25996236, 0.49374128, 0.61583621, 0.54552425, 0.62481345, 0.37726184, 0.55573486, 0.31997297, 0.38252179, 0.4798453 , 0.34475894, 0.70821782, 0.58723751, 0.30732769, 0.67060583, 0.35883077, 0.49432155, 0.63257589, 0.4746551 , 0.20712515, 0.34318549, 0.24861632, 0.40486138, 0.50773623, 0.6546179 , 0.41355075, 0.62549141, 0.43162361, 0.41401631, 0.50939969, 0.41119142, 0.45187448, 0.44351145, 0.55291979, 0.51731651, 0.52105935, 0.39136731, 0.30310754, 0.49255876, 0.45572459, 0.52494839, 0.33498898, 0.3144463 , 0.35948517, 0.29414201, 0.49625346, 0.4122108 , 0.57120471, 0.5987473 , 0.52627839, 0.79705665, 0.55928724, 0.4486908 , 0.45064562, 0.78453   , 0.65894434, 0.62606125, 0.35341933, 0.1144822 , 0.52058771, 0.48241874, 0.44551283, 0.64315633, 0.51214983, 0.38359794, 0.56489797, 0.44712031, 0.75022786, 0.49156776, 0.33048625, 0.79940974, 0.74946317, 0.38369986, 0.5302718 , 0.56759417, 0.57825614, 0.65246153, 0.31143474, 0.45214586, 0.46799504, 0.78050031, 0.36913631, 0.53373014, 0.82305238, 0.48862099, 0.36438137, 0.57922633, 0.41646953, 0.37273844, 0.39007451, 0.39836807, 0.14841261, 0.61327033, 0.66875887, 0.73502149, 0.45745652, 0.42192083, 0.46513111, 0.45418976, 0.37571297, 0.41068768, 0.5204207 , 0.59348836, 0.46156973, 0.64106824, 0.47531034, 0.45544494, 0.30021247, 0.56549417, 0.64254133, 0.79328525, 0.48902048, 0.38219606, 0.36521556, 0.65661227, 0.53675591, 0.4025876 , 0.55852129, 0.47575553, 0.4415518 , 0.35439118, 0.66019495, 0.59856318, 0.60701458, 0.56189854, 0.30719616, 0.40811194, 0.47623593, 0.79035699, 0.27784137, 0.53227073, 0.38980902, 0.5393956 , 0.71567933, 0.68997945, 0.39234727, 0.30167819, 0.66736201, 0.40535561, 0.57947614, 0.22358024, 0.28642175, 0.49378006, 0.43270344, 0.43663362, 0.67143985, 0.54729768, 0.61067125, 0.850662  , 0.21998077, 0.60404265, 0.20570871, 0.29620495, 0.53113463, 0.75943656, 0.53340705, 0.56367341, 1.05852851, 0.58699365, 0.55708933, 0.42770893, 0.63057458, 0.46624258, 0.35312807, 0.62739264, 0.72560502, 0.42181763, 0.64084169, 0.7895808 , 0.3345453 , 0.48617793, 0.3364872 , 0.56934345, 0.59753084, 0.57877394, 0.43379878, 0.62668416, 0.23533247, 0.17864523, 0.59186135, 0.3699303 , 0.46296373, 0.65727413, 0.54072042, 0.38174847, 0.27108012, 0.46935057, 0.682528  , 0.64984851, 0.45516901, 0.60251771, 0.42190568, 0.50217704, 0.63391551, 0.49945264, 0.51196726, 0.47544762, 0.53972717, 0.59143159, 0.48619282, 0.48911128, 0.41052774, 0.46348714, 0.1273272 , 0.59290562, 0.36932705, 0.3859974 , 0.45323001, 0.69940922, 0.11706349, 0.45047762, 0.57712164, 0.35027231, 0.65730514, 0.42432276, 0.63669957, 0.41655229, 0.74747243, 0.45330534, 0.56557449, 0.17116886, 0.51316985, 0.50161203, 0.4404198 , 0.61604064, 0.47152176, 0.52587154, 0.48142779, 0.61283785, 0.59740059, 0.50514449, 0.45842312, 0.59722182, 0.46923603, 0.45916802, 0.63621121, 0.57623313, 0.57514659, 0.49858844, 0.53234598, 0.48230869, 0.84355054])

# to run the full simulation uncomment the following line to fit the model for every dataset and not just for the first dataset
#for i_rep in range(n_rep):
In [16]: for i_rep in range(1):
....:     (x, y, d) = data[i_rep]
....:     this_theta = est_ols(y, np.column_stack((d, x)))
....:     print(np.abs(theta_ols[i_rep] - this_theta))
....:     theta_ols[i_rep] = this_theta
....:
3.213969090865021e-10

In [17]: ax = sns.kdeplot(theta_ols, shade=True, color=colors[0])

In [18]: ax.axvline(0.5, color='k', label='True theta');

library(ggplot2)

est_ols = function(df) {
ols = stats::lm(y ~ 1 +., df)
theta = coef(ols)["d"]
return(theta)
}

# to speed up the illustration we hard-code the simulation results
theta_ols = c(0.607697047, 0.517030100, 0.589652643, 0.472635435, 0.664240859, 0.493210806, 0.539281370, 0.314045648, 0.610190305, 0.381273649, 0.449613657, 0.591410296, 0.475418067, 0.453225281, 0.626975360, 0.133881932, 0.688402884, 0.584272744, 0.400995390, 0.508011546, 0.652163232, 0.450246441, 0.481371297, 0.387925003, 0.339319744, 0.455211505, 0.393108971, 0.508044365, 0.570043632, 0.504138457, 0.649995372, 0.573181791, 0.450204168, 0.580155929, 0.559124755, 0.319166483, 0.619605243, 0.559989398, 0.497594152, 0.516462187, 0.261960663, 0.204280903, 0.669016644, 0.250055866, 0.647017006, 0.700356932, 0.478880881, 0.490902426, 0.576694200, 0.497799178, 0.654611430, 0.761003814, 0.527117173, 0.576149401, 0.268216840, 0.427497472, 0.558284193, 0.465285351, 0.367369535, 0.358950841, 0.353317416, 0.551121683, 0.476083529, 0.349086468, 0.556019829, 0.242773180, 0.462672952, 0.661802570, 0.652935787, 0.476783262, 0.550197421, 0.605953210, 0.698217648, 0.516145147, 0.474546204, 0.592796175, 0.595952626, 0.499719871, 0.470920166, 0.635882392, 0.516079626, 0.452397659, 0.714190650, 0.215255273, 0.347379010, 0.583471048, 0.524819957, 0.608911706, 0.423763166, 0.291108689, 0.607837973, 0.541538378, 0.632318639, 0.528820038, 0.619982779, 0.570696185, 0.245974669, 0.411828619, 0.595350425, 0.547821041, 0.726634349, 0.395092155, 0.470682777, 0.246821370, 0.570568009, 0.451889824, 0.831386706, 0.474180783, 0.685116161, 0.615667439, 0.412768647, 0.420461956, 0.268491533, 0.431516537, 0.352601742, 0.592253366, 0.368416831, 0.374213036, 0.490669696, 0.374423707, 0.769901750, 0.513044163, 0.439884666, 0.560922008, 0.629981460, 0.321926767, 0.398217909, 0.444816873, 0.321365131, 0.272833380, 0.093068976, 0.445214485, 0.298279565, 0.748386201, 0.154777280, 0.617161610, 0.318668352, 0.654297770, 0.325757680, 0.343033384, 0.643452659, 0.429671013, 0.435013955, 0.507684008, 0.477399282, 0.596309464, 0.785548288, 0.727160279, 0.327009532, 0.320665984, 0.240790080, 0.517340592, 0.429096939, 0.358122017, 0.446044994, 0.581608198, 0.390101013, 0.500437297, 0.594003137, 0.475476934, 0.525995116, 0.631689118, 0.707119548, 0.462592407, 0.542751202, 0.390147315, 0.511128783, 0.415721616, 0.493532013, 0.508752438, 0.280580913, 0.633977202, 0.456987706, 0.263429085, 0.427346822, 0.631988342, 0.529404022, 0.356613732, 0.544055219, 0.486378192, 0.577250067, 0.308349783, 0.434198591, 0.759832976, 0.826757693, 0.589255696, 0.372214909, 0.634219883, 0.720108064, 0.556854714, 0.550057225, 0.275968020, 0.577875894, 0.384650579, 0.642032161, 0.470298911, 0.495913354, 0.553910651, 0.517680868, 0.438144890, 0.325649166, 0.742172310, 0.513262336, 0.372671139, 0.715375153, 0.455077113, 0.653043284, 0.491809178, 0.328426865, 0.463785639, 0.339955421, 0.630678263, 0.533009719, 0.481001903, 0.529705874, 0.543791494, 0.496534577, 0.192647347, 0.145075102, 0.632269149, 0.338828882, 0.443917462, 0.295798356, 0.323562898, 0.525703243, 0.477445901, 0.550358089, 0.468644795, 0.414003964, 0.464400822, 0.180006584, 0.267829894, 0.537841890, 0.573499595, 0.401085692, 0.590516135, 0.575580475, 0.265872644, 0.527654550, 0.638358999, 0.545591043, 0.402938792, 0.376629718, 0.249059834, 0.639295634, 0.533255230, 0.772053877, 0.556304177, 0.463626961, 0.581361233,
0.620637149, 0.333450659, 0.582466901, 0.421818439, 0.396012165, 0.384697314, 0.453416556, 0.366778974, 0.340375550, 0.267581312, 0.670849391, 0.727770265, 0.772749482, 0.438892589, 0.474494838, 0.370620526, 0.459266969, 0.747379921, 0.609076462, 0.738891475, 0.353488457, 0.545334880, 0.448778578, 0.703513317, 0.472304254, 0.315757951, 0.483654087, 0.454779120, 0.706079914, 0.670826331, 0.416380101, 0.604091197, 0.329262944, 0.720038419, 0.237782551, 0.496501223, 0.431067052, 0.428069445, 0.370304254, 0.480437628, 0.529264620, 0.495586792, 0.592154256, 0.669026555, 0.325348666, 0.527771797, 0.532025341, 0.438769788, 0.551898543, 0.463978466, 0.600295300, 0.531212791, 0.512854079, 0.627400905, 0.519508323, 0.458166808, 0.439592332, 0.516245391, 0.566304094, 0.592501171, 0.558243707, 0.268560020, 0.565398339, 0.618714551, 0.508838195, 0.660381899, 0.445473556, 0.435312543, 0.512301291, 0.257373982, 0.542377362, 0.427538306, 0.573145587, 0.458964794, 0.598656675, 0.681245473, 0.618432741, 0.379811057, 0.794911453, 0.331247062, 0.471548075, 0.500438701, 0.464223659, 0.504887335, 0.490951199, 0.751013297, 0.527585367, 0.551770873, 0.364502446, 0.569470880, 0.473540560, 0.823152096, 0.468991845, 0.457594544, 0.513934163, 0.519542386, 0.622697949, 0.608237799, 0.420012253, 0.559972124, 0.405123891, 0.445030586, 0.536857451, 0.433109716, 0.765927280, 0.751127949, 0.368974915, 0.567233309, 0.626027630, 0.386810086, 0.524774302, 0.442167054, 0.450210959, 0.713608885, 0.462550932, 0.659955300, 0.689413483, 0.382259590, 0.580821434, 0.282600974, 0.242440264, 0.386691118, 0.505693034, 0.152611856, 0.398385325, 0.473344934, 0.191897966, 0.316695260, 0.629784515, 0.274051878, 0.655870346, 0.535790488, 0.548050952, 0.632355263, 0.304193602, 0.605364088, 0.313958760, 0.614444992, 0.311886064, 0.397354327, 0.348304983, 0.692821996, 0.548204787, 0.466375294, 0.384260162, 0.598834776, 0.131651827, 0.775434060, 0.507745382, 0.495182858, 0.624167391, 0.421378417, 0.636080402, 0.396019932, 0.450640880, 0.444656038, 0.419600313, 0.531778187, 0.764677617, 0.359989103, 0.440272838, 0.420530515, 0.701544603, 0.582921387, 0.711803958, 0.644888211, 0.405292471, 0.710975237, 0.540956133, 0.403959810, 0.487418324, 0.461411210, 0.206262550, 0.555388326, 0.624785136, 0.614971417, 0.303040224, 0.369155344, 0.598749138, 0.613769413, 0.624180808, 0.743259561, 0.623789919, 0.495058489, 0.757678344, 0.545722785, 0.271155910, 0.470625056, 0.314716811, 0.392130282, 0.616851404, 0.480906333, 0.465538674, 0.786489402, 0.508033589, 0.380625856, 0.507087669, 0.457253001, 0.413843144, 0.307455347, 0.657527799, 0.601619008, 0.354791429, 0.397159635, 0.547725875, 0.591647021, 0.371593053, 0.312816249, 0.295159359, 0.613657807, 0.667318196, 0.573041019, 0.627212476, 0.594918699, 0.522951574, 0.578996755, 0.411323834, 0.696456006, 0.902523993, 0.504887743, 0.388866278, 0.542039989, 0.641256381, 0.067813714, 0.453649751, 0.553396858, 0.485506701, 0.439271561, 0.783773120, 0.655390205, 0.284032005, 0.394359328, 0.567734225, 0.411265874, 0.512826534, 0.667667737, 0.494348457, 0.754588872, 0.524358107, 0.371814862, 0.609810028, 0.398656232, 0.427543064, 0.412278778, 0.466612248, 0.508781321, 0.594168362, 0.763906439, 0.435792328, 0.627318604,
0.40128283, 0.56432264, 0.74302951, 0.64531519, 0.43428408, 0.54253650, 0.68675909, 0.54610904, 0.31143410, 0.66361247, 0.32512856, 0.66572909, 0.40574241, 0.42634426, 0.53028845, 0.34059186, 0.44227985, 0.32888353, 0.38161937, 0.61764474, 0.69864362, 0.63056189, 0.51969777, 0.58412973, 0.36162152, 0.43894783, 0.55070409, 0.61962881, 0.46346635, 0.44584711, 0.60291852, 0.54722201, 0.32348962, 0.40724896, 0.60074412, 0.42378812, 0.53906840, 0.45414918, 0.59115562, 0.40338927, 0.50182640, 0.48084020, 0.90263159, 0.60790853, 0.43655100, 0.55461189, 0.36397835, 0.58646397, 0.58908656, 0.59499250, 0.41971334, 0.55403415, 0.48406753, 0.51324153, 0.43795068, 0.34147378, 0.65615424, 0.45282946, 0.62798226, 0.68307839, 0.34736619, 0.27418656, 0.42805168, 0.56468975, 0.60789152, 0.59062508, 0.49657063, 0.25934902, 0.46669355, 0.37904474, 0.78088127, 0.27779921, 0.41089186, 0.50652618, 0.41487435, 0.42474508, 0.28787443, 0.68609987, 0.40486436, 0.58286324, 0.69899550, 0.69881034, 0.25414322, 0.50056500, 0.51191289, 0.64195381, 0.38464824, 0.67050520, 0.79119204, 0.66121900, 0.66231864, 0.49102378, 0.52755161, 0.37539878, 0.49396307, 0.56514515, 0.64396712, 0.58492056, 0.51141379, 0.46374323, 0.61322274, 0.80605473, 0.19625903, 0.59275028, 0.59315605, 0.51476527, 0.55586894, 0.62692145, 0.85075221, 0.22599649, 0.59567280, 0.42571674, 0.57336632, 0.47226078, 0.42181200, 0.61536354, 0.57852012, 0.56960488, 0.67248512, 0.57019950, 0.72583961, 0.39351170, 0.45864071, 0.43103856, 0.50799773, 0.43716181, 0.62919779, 0.48415089, 0.65312876, 0.65647993, 0.46407403, 0.33089356, 0.37438204, 0.37497296, 0.68611434, 0.44876905, 0.18054140, 0.26400354, 0.47987279, 0.68171662, 0.51093778, 0.59986890, 0.30665047, 0.55657833, 0.28716139, 0.50404870, 0.57869547, 0.71382251, 0.23780666, 0.74193622, 0.46573069, 0.61869736, 0.50879341, 0.49846749, 0.52458970, 0.60034101, 0.43890792, 0.35354176, 0.62410794, 0.47428766, 0.67168489, 0.49854470, 0.58477878, 0.49254631, 0.53694318, 0.51833287, 0.61090065, 0.55870688, 0.32501375, 0.48465602, 0.60557801, 0.65122122, 0.58698711, 0.62996355, 0.79432546, 0.29601187, 0.37071731, 0.13453090, 0.73666178, 0.60634991, 0.43765982, 0.47549840, 0.47018984, 0.71373865, 0.62206020, 0.50556747, 0.32655121, 0.73705585, 0.80461499, 0.43967304, 0.51218814, 0.52506201, 0.74933665, 0.31768015, 0.41512105, 0.36382194, 0.31293697, 0.22764901, 0.45876455, 0.58077527, 0.34656751, 0.50245026, 0.54243241, 0.29990733, 0.40296793, 0.51772611, 0.53623182, 0.65568813, 0.41566544, 0.52366193, 0.68709611, 0.49724011, 0.63823942, 0.57515874, 0.62556556, 0.54638853, 0.24648101, 0.45804953, 0.37131632, 0.60435690, 0.47646836, 0.51087791, 0.60959009, 0.41817240, 0.30132056, 0.75619161, 0.51393051, 0.51373308, 0.71556140, 0.30710997, 0.56009148, 0.41724733, 0.45965438, 0.56950917, 0.60834196, 0.33933807, 0.63838966, 0.83730647, 0.64062354, 0.41755781, 0.41221805, 0.62107756, 0.38217892, 0.65674740, 0.62836579, 0.41161122, 0.45185653, 0.45227430, 0.76045195, 0.41824912,
0.4810422212,  0.6320157397,  0.5921241513,  0.5766938495,  0.3247115737,  0.2703896827,  0.4185879263,  0.7592021572,  0.5332725283,  0.6542020920,  0.6328260609,  0.4924104186,  0.4036824886,  0.5118722903,  0.3665288504,  0.3426986470,  0.4015860810,  0.5457729851,  0.5617826399,  0.2457019273,  0.6232354343,  0.6198344522,  0.3336892970,  0.1571112376,  0.5302578543,  0.5896195835,  0.5045548852,  0.1038594426,  0.4827105478,  0.4625021003,  0.6622099645,  0.3906701212,  0.7855141344,  0.4739854013,  0.5974381160,  0.4956047962,  0.3884149227,  0.4735520726,  0.5569370070,  0.3885771241,  0.6921680637,  0.5307979053,  0.4478790041,  0.2521418794,  0.3653833498,  0.4644757108,  0.4920318549,  0.4954566442,  0.5570176214,  0.6502358567,  0.4753184434,  0.4594743395,  0.5482583307,  0.6277798133,  0.6981686206,  0.6187655987,  0.5965462283,  0.5297084261,  0.4279360901,  0.1817626048,  0.2111365738,  0.3798967352,  0.3077189146,  0.5074858352,  0.3020465101,  0.5585563587,  0.4669719417,  0.6258874623,  0.4788181254,  0.6031830563,  0.4192175248,  0.8315681795,  0.7514063990,  0.5105236249,  0.3613152818,  0.6270093033,  0.3523742755,  0.6425440918,  0.3457825041,  0.5677974519,  0.4554885768,  0.3853195932,  0.6944759319,  0.7279363416,  0.6307611619,  0.7266577792,  0.4691833321,  0.4364350803,  0.2740016735,  0.3683804337,  0.3656010469,  0.4847634576,  0.2635796128,  0.4708807769,  0.4251331735,  0.5961937080,  0.7359779525,  0.3960925525,  0.7544216824,  0.4553743674,  0.6277900183,  0.5604000678,  0.4123944519,  0.6697466795,  0.4542655060,  0.6726582582,  0.5178095934,  0.5336656003,  0.7294123494,  0.5737851442,  0.4375373221,  0.5169854604,  0.5647916007,  0.5281703047,  0.4947867579,  0.6033878160,  0.4348261019,  0.6622941134,  0.5181854618,  0.4387885592,  0.5840565870,  0.7078765595,  0.4293114362,  0.5289573078,  0.6357464673,  0.4019810076,  0.4917680291,  0.5141892462,  0.5742675024,  0.6187602570,  0.4924936579,  0.4631088728,  0.2970657923,  0.6183528546,  0.5926991307,  0.3422548365,  0.4630764069,  0.3745998673,  0.5779962672,  0.5943075376,  0.4863586326,  0.5702442510,  0.8110391970,  0.4637084427,  0.7044627148,  0.5141194878,  0.4610210390,  0.4283283243,  0.2124464880,  0.3241674624,  0.8986303116,  0.6650904154,  0.4358467530,  0.3623206466,  0.3086154653,  0.4073734322,  0.3954349869,  0.7140152024,  0.5475629076,  0.3772126235,  0.6549230244,  0.7222120900,  0.6293562288,  0.7074974891,  0.3862122773,  0.5190522868,  0.5086677172,  0.4566653024,  0.6198970108,  0.5534357746,  0.6403650132,  0.4776773409,  0.7993807655,  0.3316596983,  0.5347903151, -0.0073145945,  0.7505758880,  0.7721696602,  0.4786098286,  0.7733355763,  0.4292071198,  0.5524794380,  0.4738381907,  0.4532365578,  0.3864576996,  0.3154620027,  0.4125880055,  0.7336049906,  0.7191266612,  0.4174040603,  0.3966765475,  0.4200732061,  0.5158521385,  0.5179938216,  0.6002159767,  0.6655745967,  0.3359239334,  0.2815784591,  0.3850891300,  0.7365577578,  0.3621555499,  0.6444221888,  0.5666878213,  0.5194972086,  0.7835955228,  0.8471366245,  0.6936365628,  0.2793957788,  0.5781979405,  0.3851868684,  0.5080473573,  0.5638570327,  0.2292095223,  0.5795689538,  0.4331280258,  0.4737607122,  0.5558744047,  0.6336470554,  0.5559408316,  0.4499369967,  0.6496618608,  0.6065969525,  0.3587967152,  0.6153089390,  0.5826552712,  0.6817230843,  0.4902681535,  0.4304772497,  0.3568752494,  0.4414055507,  0.5421596269,  0.6616140050,  0.5645489367,  0.4600724670,  0.5220926673,  0.2790394247,  0.7390774505,  0.3820212110,  0.5281986424,  0.6567182815,  0.4522734618,  0.3908002136,  0.7182619596,  0.6261977128,  0.2436503360,  0.3906712092,  0.4287648988,  0.5611551354,  0.7583334768,  0.3711041209)
# to run the full simulation uncomment the following line to fit the model for every dataset and not just for the first dataset
#for (i_rep in seq_len(n_rep)) {
for (i_rep in seq_len(1)) {
df = data[[i_rep]]
this_theta = est_ols(df)
print(abs(theta_ols[i_rep] - this_theta))
theta_ols[i_rep] = this_theta
}

g_ols = ggplot(data.frame(theta_ols), aes(x = theta_ols)) +
geom_density(fill = "dark blue", alpha = 0.3, color = "dark blue") +
geom_vline(aes(xintercept = alpha), col = "black") +
xlim(c(0.08, 0.75)) + xlab("") + ylab("") + theme_minimal()
g_ols

          d
2.20214e-10


## 1.3. Regularization bias in simple ML-approaches¶

Naive inference that is based on a direct application of machine learning methods to estimate the causal parameter, $$\theta$$, is generally invalid. The use of machine learning methods introduces a bias that arises due to regularization. A simple ML approach is given by randomly splitting the sample into two parts. On the auxiliary sample indexed by $$i \in I^C$$ the nuisance function $$g_0(X)$$ is estimated with an ML method, for example a random forest learner. Given the estimate $$\hat{g}_0(X)$$, the final estimate of $$\theta$$ is obtained as ($$n=N/2$$) using the other half of observations indexed with $$i \in I$$

$\hat{\theta} = \left(\frac{1}{n} \sum_{i\in I} D_i^2\right)^{-1} \frac{1}{n} \sum_{i\in I} D_i (Y_i - \hat{g}_0(X_i)).$
In [19]: def non_orth_score(y, d, g_hat, m_hat, smpls):
....:     u_hat = y - g_hat
....:     psi_a = -np.multiply(d, d)
....:     psi_b = np.multiply(d, u_hat)
....:     return psi_a, psi_b
....:

In [20]: from doubleml import DoubleMLData

In [21]: from doubleml import DoubleMLPLR

In [22]: from sklearn.ensemble import RandomForestRegressor

In [23]: from sklearn.base import clone

In [24]: import numpy as np

In [25]: np.random.seed(1111)

In [26]: learner = RandomForestRegressor(n_estimators=100, max_features=n_vars, max_depth=5, min_samples_leaf=2)

In [27]: ml_m = clone(learner)

In [28]: ml_g = clone(learner)

# to speed up the illustration we hard-code the simulation results
In [29]: theta_nonorth = np.array([0.46220293, 0.36751016, 0.33633589, 0.40576358, 0.41204023, 0.44099791, 0.29077441, 0.51049579, 0.40872897, 0.34517717, 0.56135355, 0.20250422, 0.30014152, 0.3121669 , 0.48135903, 0.4250507 , 0.37715598, 0.37129675, 0.48191319, 0.45303646, 0.23189062, 0.47421963, 0.46100559, 0.51690426, 0.37912305, 0.3164967 , 0.5018225 , 0.34575523, 0.40272864, 0.31201177, 0.2868938 , 0.43818704, 0.5791864 , 0.37342275, 0.61239568, 0.35811336, 0.29545835, 0.49732933, 0.40667421, 0.32322828, 0.51692541, 0.2692527 , 0.14121754, 0.33655586, 0.33198124, 0.25389419, 0.38976939, 0.32516571, 0.24778959, 0.45216823, 0.62536581, 0.33626136, 0.48605982, 0.43367237, 0.22609483, 0.45695586, 0.51988   , 0.36181456, 0.3017518 , 0.35912742, 0.33766141, 0.28139109, 0.33423182, 0.28849494, 0.35598731, 0.30291489, 0.26696673, 0.3965472 , 0.30285673, 0.38740067, 0.38927095, 0.37276621, 0.24221399, 0.2478911 , 0.19369063, 0.49629465, 0.24420607, 0.46295666, 0.38999345, 0.28754147, 0.40750691, 0.3474621 , 0.2714744 , 0.40731088, 0.40115083, 0.3567057 , 0.39989027, 0.49022217, 0.25400176, 0.53195003, 0.37218689, 0.47224868, 0.29372151, 0.29748552, 0.15382988, 0.31676935, 0.4371993 , 0.58477685, 0.47081611, 0.24718853, 0.36163077, 0.24769679, 0.35941911, 0.3296207 , 0.30717881, 0.26779407, 0.3010835 , 0.31544093, 0.53523355, 0.32454461, 0.3090798 , 0.59057928, 0.40864869, 0.30559666, 0.51896589, 0.2602647 , 0.44938393, 0.40882983, 0.39964357, 0.18228496, 0.26144559, 0.40876431, 0.27486033, 0.41148144, 0.40479421, 0.37690703, 0.43214482, 0.41183592, 0.30884357, 0.4395378 , 0.30733564, 0.28518014, 0.51267318, 0.418407  , 0.21566437, 0.38225971, 0.44524558, 0.44356552, 0.30699452, 0.28843287, 0.29729574, 0.40998509, 0.3152678 , 0.37034559, 0.32783926, 0.21489544, 0.30194505, 0.33168137, 0.57711369, 0.42932541, 0.30693366, 0.47730183, 0.31486424, 0.49108048, 0.37679733, 0.13837656, 0.38903388, 0.27232008, 0.25654145, 0.30411392, 0.32742528, 0.18762933, 0.35556439, 0.41665679, 0.36389463, 0.28492621, 0.22649755, 0.36533554, 0.32661319, 0.36891703, 0.5346196 , 0.32402164, 0.41344347, 0.43807866, 0.44701831, 0.28145691, 0.20700588, 0.32530295, 0.43505573, 0.56001566, 0.35948589, 0.49295628, 0.48489552, 0.32220793, 0.38433705, 0.32862821, 0.26534293, 0.38470655, 0.5317061 , 0.3155212 , 0.41230886, 0.31880469, 0.38930306, 0.30419097, 0.28164819, 0.44003141, 0.52568488, 0.30156178, 0.42740164, 0.25484798, 0.52973602, 0.37305963, 0.50358078, 0.40041206, 0.40130874, 0.32731488, 0.12814492, 0.54048524, 0.30212646, 0.29944921, 0.46193322, 0.48621964, 0.36634665, 0.40813239, 0.54298408, 0.22190248, 0.39525592, 0.69387251, 0.20593708, 0.28880902, 0.42479694, 0.31044302, 0.22034116, 0.31740497, 0.24030425, 0.56976883, 0.3957697 , 0.34617178, 0.29688818, 0.28557196, 0.31972769, 0.4681434 , 0.13060227, 0.28188695, 0.40322242, 0.40874094, 0.39828389, 0.35414176, 0.29122915, 0.54909401, 0.44888364, 0.50358742, 0.25438398, 0.23197   , 0.51684221, 0.48795009, 0.27609755, 0.63220386, 0.26865073, 0.50212524, 0.26094088, 0.3441297 , 0.4248767 , 0.3917966 , 0.28339179, 0.49063038, 0.23897118, 0.23529262, 0.44717751, 0.32795002, 0.32401678, 0.38156285, 0.34710818, 0.27102277, 0.28859347, 0.34379081, 0.37989127, 0.23334585, 0.3858822 , 0.24988656, 0.32844847, 0.17907128, 0.4724534 , 0.55511725, 0.49210929, 0.51413831, 0.48496465, 0.33768297, 0.45732113, 0.16237709, 0.31259887, 0.24780099, 0.44087839, 0.52862188, 0.40255284, 0.43976953, 0.39427105, 0.43236244, 0.3808625 , 0.27766508, 0.49947025, 0.38035409, 0.5427271 , 0.31245137, 0.26999138, 0.37705422, 0.24665438, 0.15804536, 0.23302156, 0.25452638, 0.20015759, 0.34178481, 0.42732665, 0.26793923, 0.432119  , 0.36664793, 0.28412842, 0.31266903, 0.37151239, 0.41072197, 0.25914919, 0.34766146, 0.35755162, 0.47770275, 0.57342129, 0.33459541, 0.26022411, 0.32527293, 0.48254574, 0.46709962, 0.32910331, 0.29890511, 0.30003679, 0.35706126, 0.67180482, 0.54421382, 0.51935982, 0.40963279, 0.48963057, 0.34349731, 0.45373368, 0.52089024, 0.18370033, 0.51526784, 0.25555252, 0.13523622, 0.41918706, 0.38274286, 0.25636761, 0.34090336, 0.38543472, 0.3500451 , 0.44221631, 0.47848301, 0.56810765, 0.28853275, 0.31266286, 0.08315981, 0.37323175, 0.37050492, 0.37132835, 0.39176757, 0.28487848, 0.34811582, 0.31341682, 0.25741468, 0.45567148, 0.42655754, 0.43384906, 0.4575956 , 0.41901164, 0.27669954, 0.28489968, 0.57725093, 0.49211554, 0.36421352, 0.42156277, 0.38157618, 0.22128394, 0.22988936, 0.37800291, 0.32438333, 0.42074574, 0.50153677, 0.4335348 , 0.39228647, 0.41168389, 0.30185482, 0.54737502, 0.33114702, 0.29556278, 0.24722737, 0.42596609, 0.27549308, 0.26901175, 0.36913136, 0.54180479, 0.2507623 , 0.36728341, 0.32845508, 0.37350956, 0.50933777, 0.16048549, 0.20709023, 0.33046893, 0.40475173, 0.49111455, 0.3316136 , 0.42223533, 0.301624  , 0.32762598, 0.43085753, 0.37455014, 0.38346095, 0.28269517, 0.31832102, 0.39854608, 0.53785366, 0.30459447, 0.25608585, 0.31575244, 0.51918424, 0.44011651, 0.48129789, 0.31309313, 0.38842631, 0.32047838, 0.51359531, 0.2291564 , 0.50243652, 0.30439938, 0.3842209 , 0.45754726, 0.34691952, 0.40119005, 0.42947326, 0.35732844, 0.7053636 , 0.51140155, 0.32063274, 0.34249698, 0.1874778 , 0.21482314, 0.48836059, 0.38574774, 0.34699361, 0.49846891, 0.20681247, 0.32080384, 0.33003434, 0.19575346, 0.40129487, 0.28568734, 0.38648456, 0.42028374, 0.31780298, 0.42404296, 0.42373512, 0.40568125, 0.47821789, 0.27384261, 0.23192378, 0.33682487, 0.2536611 , 0.31202691, 0.48949854, 0.45027394, 0.26699838, 0.23500005, 0.35938584, 0.32315096, 0.51400505, 0.5476829 , 0.47742672, 0.615812  , 0.17242558, 0.48755003, 0.31894441, 0.41555802, 0.30524289, 0.36586943, 0.30051922, 0.50294205, 0.37651153, 0.43522539, 0.49755294, 0.51112365, 0.25821398, 0.28890604, 0.55368788, 0.38367044, 0.268415  , 0.44027297, 0.05237823, 0.39204202, 0.36754218, 0.52114321, 0.34114925, 0.50433419, 0.50492222, 0.36294595, 0.4977941 , 0.44289807, 0.32654641, 0.40918277, 0.46906927, 0.64750542, 0.51347401, 0.30161206, 0.36895661, 0.54105561, 0.29029735, 0.51704962, 0.37911813, 0.40022072, 0.335743  , 0.40777632, 0.2378021 , 0.50441299, 0.46180594, 0.41043469, 0.40280337, 0.42049487, 0.52966804, 0.26675556, 0.54892945, 0.36067864, 0.35039717, 0.31671971, 0.41219094, 0.58326159, 0.29431086, 0.22111439, 0.27925197, 0.19789639, 0.43056906, 0.37760997, 0.35933596, 0.2671348 , 0.41115347, 0.21252915, 0.54231981, 0.41166753, 0.4394692 , 0.33965765, 0.42278529, 0.38197679, 0.43796052, 0.38463293, 0.3622347 , 0.34590094, 0.45161908, 0.42775285, 0.3514514 , 0.29216989, 0.56870671, 0.3728789 , 0.4840625 , 0.24939817, 0.51749173, 0.33212043, 0.33962697, 0.37728984, 0.51574754, 0.32186049, 0.35954392, 0.23945795, 0.51635551, 0.36124072, 0.17971819, 0.2733213 , 0.23709181, 0.28921415, 0.37950353, 0.35832946, 0.33919053, 0.4421467 , 0.44453859, 0.3084489 , 0.40711688, 0.18442461, 0.33353401, 0.27882742, 0.27196402, 0.27871357, 0.43660472, 0.40557353, 0.38828503, 0.29888806, 0.3077158 , 0.3003538 , 0.32188203, 0.44470064, 0.43177137, 0.2478142 , 0.40921662, 0.34603449, 0.40827106, 0.51307184, 0.31155588, 0.2422153 , 0.3897607 , 0.1496668 , 0.54856908, 0.30244926, 0.51141468, 0.41901821, 0.44441977, 0.3978117 , 0.3297629 , 0.40634361, 0.62277346, 0.22786149, 0.14830198, 0.37737936, 0.34199748, 0.25706684, 0.54490589, 0.28239921, 0.46276807, 0.45896931, 0.24764173, 0.4424539 , 0.45002468, 0.27890536, 0.32633905, 0.45257513, 0.41932199, 0.19561978, 0.55144693, 0.51333353, 0.45842698, 0.56147671, 0.36887492, 0.38624545, 0.26848941, 0.30304273, 0.32412876, 0.26750913, 0.35213321, 0.26239338, 0.27360748, 0.45712846, 0.23944032, 0.28400179, 0.24137251, 0.33875325, 0.43673468, 0.51051326, 0.39558491, 0.33674728, 0.31155873, 0.35082617, 0.3450042 , 0.2068982 , 0.42893076, 0.53609975, 0.43484156, 0.39190293, 0.40235957, 0.43536865, 0.13199344, 0.37234453, 0.36487717, 0.26998345, 0.3577768 , 0.21139084, 0.29233936, 0.31770874, 0.33413338, 0.41943277, 0.42868319, 0.44901983, 0.39465012, 0.43047094, 0.40858499, 0.32071093, 0.41479505, 0.43548872, 0.64042886, 0.44519231, 0.33936638, 0.36481828, 0.31619029, 0.33006278, 0.44194993, 0.35762225, 0.44942049, 0.47223845, 0.35970479, 0.44020136, 0.50150471, 0.45563746, 0.41909168, 0.5562186 , 0.35541362, 0.32981698, 0.3120512 , 0.51529814, 0.54700928, 0.40025523, 0.30124641, 0.52107093, 0.30245403, 0.57525517, 0.31773368, 0.27561939, 0.34122497, 0.48125339, 0.33556223, 0.34537586, 0.46446115, 0.53067435, 0.33114095, 0.04002178, 0.25309885, 0.43136239, 0.47662659, 0.30171256, 0.27291973, 0.36485227, 0.49433584, 0.53073764, 0.32822755, 0.30892213, 0.19229913, 0.49244806, 0.24253362, 0.23946686, 0.41228701, 0.23044998, 0.44281423, 0.38381084, 0.39892405, 0.25826518, 0.26965326, 0.19635469, 0.36782008, 0.32739637, 0.48639554, 0.26193771, 0.32535793, 0.42626676, 0.39908775, 0.46175108, 0.46276024, 0.19055886, 0.30795696, 0.26053898, 0.30538593, 0.39416142, 0.30411534, 0.49163854, 0.34706847, 0.25326895, 0.38240986, 0.39468779, 0.38342138, 0.43454342, 0.49144649, 0.50407511, 0.25807762, 0.12567515, 0.39743335, 0.45415779, 0.49915046, 0.41095417, 0.28133215, 0.37674896, 0.32845893, 0.424317  , 0.40249317, 0.32678539, 0.2820127 , 0.21170487, 0.28621951, 0.38618504, 0.36395987, 0.4384132 , 0.50185284, 0.38856527, 0.3355761 , 0.545829  , 0.42157608, 0.23829926, 0.35944156, 0.46892831, 0.30006905, 0.31508441, 0.13056881, 0.22983941, 0.40952014, 0.34881462, 0.51501957, 0.23724303, 0.3692855 , 0.20709138, 0.5528277 , 0.37372525, 0.44065652, 0.43773557, 0.44776214, 0.51288475, 0.43011009, 0.33546982, 0.43070946, 0.47193684, 0.26917059, 0.39031249, 0.47324386, 0.56204637, 0.37772789, 0.33920943, 0.32931146, 0.21342614, 0.39590236, 0.32683289, 0.52570395, 0.44637159, 0.49300479, 0.42811246, 0.37286832, 0.47615932, 0.29102038, 0.42614121, 0.37045208, 0.43181565, 0.39192547, 0.30357706, 0.52822227, 0.34520693, 0.15488365, 0.36314285, 0.32952075, 0.39123625, 0.45835816, 0.3492826 , 0.38845001, 0.20492465, 0.46366043, 0.45232324, 0.27062416, 0.45025608, 0.37671507, 0.44170926, 0.13905899, 0.48392469, 0.39288254, 0.37721935, 0.40392402, 0.33461498, 0.45036463, 0.29748615, 0.50968753, 0.26730238, 0.24005959, 0.38183551, 0.38616646, 0.18126938, 0.34880799, 0.28623105, 0.57572819, 0.2573925 , 0.37693313, 0.39919801, 0.4085831 , 0.4380114 , 0.22884927, 0.51024134, 0.51045657, 0.26103373, 0.21835691, 0.31725475, 0.53588833, 0.32612813, 0.22062382, 0.29764379, 0.3413454 , 0.49996528, 0.53420137, 0.3245859 , 0.41881072, 0.519687  , 0.25862658, 0.3193657 , 0.41183977, 0.39935424, 0.21932385, 0.28277402, 0.35613808, 0.51934429, 0.41845812, 0.53898359, 0.34761971, 0.26069961, 0.33369592, 0.37569057, 0.28494363, 0.50120447, 0.51826334, 0.33040705, 0.31342893, 0.37243041, 0.42314929, 0.22700338, 0.27982161, 0.35893721, 0.64832742, 0.36727074, 0.44908435, 0.17402012, 0.38247934, 0.46182565, 0.5271202 , 0.35202818, 0.30764646, 0.06996173, 0.48846568, 0.44200333, 0.31534363, 0.53015202, 0.37123406, 0.3615767 , 0.51298256, 0.50919104, 0.37256098, 0.3157192 , 0.31559631, 0.26691721, 0.44900753, 0.5329662 , 0.28359587, 0.4504041 , 0.25494302, 0.36263564, 0.43255345, 0.27407261, 0.39549419, 0.15953733, 0.30779378, 0.60872717, 0.50432529, 0.48100406, 0.46652201, 0.32864931, 0.20978906, 0.29146115, 0.53688804, 0.35460348, 0.31504874, 0.50239211, 0.32996978, 0.52825463, 0.36588116, 0.39051713, 0.35607859, 0.33992489, 0.25465816, 0.45846204, 0.33398612, 0.3052152 , 0.28117745, 0.29579359, 0.51187272, 0.3966739 , 0.41773266, 0.41882469, 0.15269325, 0.34683192, 0.21951749, 0.38644923, 0.25555259, 0.52044651, 0.39453725, 0.20224927, 0.28396396, 0.31936779, 0.25661765, 0.16838062, 0.38081762, 0.37654759, 0.2889413 , 0.40671267, 0.38476536, 0.21671325, 0.58133803, 0.28200238, 0.34528347, 0.1572706 , 0.41332857, 0.3043431 , 0.35589781, 0.34931822, 0.44463928, 0.21289216, 0.30333574, 0.3463518 , 0.37693294, 0.21480485, 0.34142576, 0.43366436, 0.4640764 , 0.42274242, 0.42184705, 0.4473984 , 0.41991344, 0.502675  , 0.38368096, 0.2441497 , 0.52133242])

# to run the full simulation uncomment the following line to fit the model for every dataset and not just for the first dataset
#for i_rep in range(n_rep):
In [30]: for i_rep in range(1):
....:     (x, y, d) = data[i_rep]
....:     obj_dml_data = DoubleMLData.from_arrays(x, y, d)
....:     obj_dml_plr_nonorth = DoubleMLPLR(obj_dml_data,
....:                                       ml_m, ml_g,
....:                                       n_folds=2,
....:                                       apply_cross_fitting=False,
....:                                       score=non_orth_score)
....:     obj_dml_plr_nonorth.fit()
....:     this_theta = obj_dml_plr_nonorth.coef[0]
....:     print(np.abs(theta_nonorth[i_rep] - this_theta))
....:     theta_nonorth[i_rep] = this_theta
....:
0.003633624559268245

In [31]: ax = sns.kdeplot(theta_nonorth, shade=True, color=colors[1])

In [32]: ax.axvline(0.5, color='k', label='True theta');

non_orth_score = function(y, d, g_hat, m_hat, smpls) {
u_hat = y - g_hat
psi_a = -1*d*d
psi_b = d*u_hat
psis = list(psi_a = psi_a, psi_b = psi_b)
return(psis)
}

library(mlr3)
library(mlr3learners)
library(data.table)
lgr::get_logger("mlr3")$set_threshold("warn") set.seed(1111) learner = lrn("regr.ranger", num.trees=100, mtry=n_vars, min.node.size=2, max.depth=5) ml_m = learner$clone()
ml_g = learner$clone() # to speed up the illustration we hard-code the simulation results theta_nonorth = theta_nonorth = c(0.335140973, 0.419119658, 0.543364882, 0.465531595, 0.489423796, 0.580690927, 0.446899757, 0.272292296, 0.240826014, 0.365608710, 0.268312379, 0.354473099, 0.385506555, 0.615247222, 0.463371670, 0.410046844, 0.544959995, 0.489229912, 0.457527626, 0.614892818, 0.338072758, 0.394127768, 0.276357084, 0.433449188, 0.249084049, 0.528947106, 0.451120107, 0.327303152, 0.546307385, 0.395934562, 0.301388718, 0.239982004, 0.530202292, 0.169856178, 0.322835980, 0.065608845, 0.154554350, 0.430897491, 0.424487779, 0.324143676, 0.389261597, 0.272135966, 0.351868551, 0.256792473, 0.504886282, 0.474738538, 0.384538245, 0.483251852, 0.363105174, 0.216916587, 0.294053708, 0.479088598, 0.269570359, 0.377226775, 0.437593329, 0.225371600, 0.470220118, 0.580390685, 0.304737611, 0.305733674, 0.354465325, 0.436580536, 0.445360886, 0.368723539, 0.157181386, 0.155664995, 0.464140152, 0.480734551, 0.362887864, 0.337695182, 0.642324202, 0.310090948, 0.349702691, 0.305153800, 0.316219058, 0.381051853, 0.320910647, 0.469491087, 0.424923136, 0.429822780, 0.275509449, 0.323183059, 0.555469839, 0.191757655, 0.236563643, 0.431688361, 0.428128667, 0.352089478, 0.099148330, 0.444152504, 0.394307625, 0.255398535, 0.438946752, 0.253313216, 0.112699448, 0.477861515, 0.458580858, 0.583866474, 0.340893368, 0.591232974, 0.346205252, 0.543514451, 0.460809329, 0.231994475, 0.284534860, 0.328665723, 0.316603063, 0.364736497, 0.326819423, 0.360088787, 0.437105307, 0.340068129, 0.437790456, 0.293644690, 0.256570227, 0.391332290, 0.379705449, 0.345781442, 0.192708760, 0.495761272, 0.357445752, 0.289328987, 0.271279462, 0.246514418, 0.321481385, 0.547511147, 0.323289625, 0.421790101, 0.361704227, 0.183561036, 0.240601590, 0.373376260, 0.418223589, 0.234931680, 0.298725451, 0.334964767, 0.445040143, 0.411404746, 0.296765746, 0.342425930, 0.439660376, 0.449429060, 0.350347796, 0.501114430, 0.141751964, 0.460308677, 0.308125041, 0.412797756, 0.348937055, 0.320860597, 0.501787986, 0.402516515, 0.307775795, 0.265250174, 0.249370199, 0.237803476, 0.490491083, 0.392297291, 0.476605166, 0.596921313, 0.329334145, 0.168666169, 0.514855555, 0.361304938, 0.458823659, 0.218943774, 0.190659814, 0.399702615, 0.496828233, 0.364189324, 0.355013117, 0.347135933, 0.303363453, 0.388743993, 0.351564875, 0.461375525, 0.304952081, 0.305842312, 0.371825913, 0.529941260, 0.369612156, 0.227549238, 0.476589099, 0.409674606, 0.648042086, 0.454762542, 0.325161264, 0.474848225, 0.238840882, 0.246974012, 0.281007912, 0.485805697, 0.349186121, 0.240147788, 0.400154659, 0.350860610, 0.321397152, 0.329311833, 0.412791960, 0.416868304, 0.365255439, 0.224506192, 0.627734333, 0.315886799, 0.187565232, 0.321475886, 0.277727795, 0.298443734, 0.490262959, 0.260242409, 0.312206710, 0.352033429, 0.330884704, 0.258199280, 0.430870900, 0.325387453, 0.366884342, 0.388394987, 0.332531847, 0.456149765, 0.480522342, 0.198383637, 0.200533915, 0.269881924, 0.339918216, 0.453854936, 0.473168756, 0.460025119, 0.284768638, 0.360963578, 0.210442939, 0.299954368, 0.389141673, 0.361433498, 0.443818155, 0.380874367, 0.315574954, 0.514698977, 0.369786833, 0.342736404, 0.496349510, 0.539101645, 0.301455561, 0.273359613, 0.470802282, 0.435284567, 0.515617885, 0.284359396, 0.311403925, 0.324005567, 0.42139206, 0.25604031, 0.36201789, 0.10846016, 0.41784835, 0.34505290, 0.35388876, 0.74150699, 0.16133690, 0.29388264, 0.47603740, 0.26673708, 0.36747984, 0.32719126, 0.42509806, 0.35977158, 0.33368013, 0.50825836, 0.43838805, 0.44730481, 0.34552361, 0.35458993, 0.23337476, 0.26513718, 0.47182588, 0.44153584, 0.47829997, 0.26140131, 0.25369408, 0.38922793, 0.15108407, 0.21647681, 0.43478364, 0.38368258, 0.36937969, 0.39530397, 0.47783846, 0.47734570, 0.23662392, 0.44670463, 0.48737923, 0.48664646, 0.37433932, 0.40340501, 0.42819694, 0.50312924, 0.26157946, 0.31198968, 0.39955912, 0.33634231, 0.33144015, 0.31181757, 0.33995580, 0.43817849, 0.30326375, 0.31102104, 0.32270600, 0.35016055, 0.40434170, 0.46299218, 0.48458540, 0.32291477, 0.38090308, 0.37289522, 0.49450332, 0.37622246, 0.26345370, 0.21200119, 0.20610146, 0.40286913, 0.30702282, 0.24507251, 0.34157918, 0.34007658, 0.32387305, 0.35049316, 0.37436835, 0.34540160, 0.48407376, 0.46762908, 0.37074398, 0.30789111, 0.35586455, 0.34952500, 0.50174516, 0.37325996, 0.27937008, 0.40428039, 0.33630248, 0.29045981, 0.36996854, 0.38606302, 0.29149853, 0.47210709, 0.51180091, 0.29678938, 0.35412552, 0.39716491, 0.28452926, 0.26660583, 0.46892162, 0.30673355, 0.45047690, 0.42939064, 0.45773370, 0.32853933, 0.46402431, 0.31983450, 0.29105138, 0.18750465, 0.32012184, 0.37348488, 0.37709558, 0.30038336, 0.35647914, 0.51807477, 0.20440505, 0.25672821, 0.27799653, 0.35506732, 0.34781610, 0.50966358, 0.35900531, 0.51515573, 0.41425917, 0.32255047, 0.43121963, 0.49713000, 0.30640585, 0.38909694, 0.29476253, 0.40916563, 0.52696848, 0.49826146, 0.38664316, 0.35355361, 0.28926478, 0.45166077, 0.35449890, 0.34640176, 0.36247956, 0.49048200, 0.30049993, 0.36595049, 0.46345042, 0.40938684, 0.43713582, 0.47727673, 0.37483511, 0.47145995, 0.31764164, 0.37748722, 0.40966739, 0.36564739, 0.68316261, 0.30395606, 0.35009298, 0.38868867, 0.42508677, 0.19270357, 0.31445686, 0.35898081, 0.54066015, 0.55762859, 0.39909256, 0.38553908, 0.42839512, 0.31360412, 0.43517347, 0.35676783, 0.31794530, 0.33347841, 0.47641606, 0.46035693, 0.35144907, 0.45348246, 0.43848581, 0.31090602, 0.42554573, 0.32571953, 0.29110146, 0.42773931, 0.47804920, 0.36678412, 0.56647980, 0.35099766, 0.42244372, 0.24605076, 0.32744448, 0.36130732, 0.24298803, 0.37705319, 0.27561302, 0.41546887, 0.42277679, 0.35905028, 0.52871705, 0.31363500, 0.35694760, 0.20090708, 0.39972094, 0.32729766, 0.38047097, 0.30324345, 0.50077595, 0.42323764, 0.45558074, 0.37947002, 0.40284181, 0.30131654, 0.35846660, 0.27345362, 0.21105024, 0.43491772, 0.27462259, 0.44970539, 0.33962746, 0.47221672, 0.38079351, 0.38404269, 0.36638005, 0.45460183, 0.41530605, 0.51348554, 0.41357115, 0.38602219, 0.39641486, 0.36338585, 0.53908472, 0.40178164, 0.37625833, 0.34680086, 0.38400230, 0.32646933, 0.43481472, 0.40057902, 0.46015932, 0.35390319, 0.49936377, 0.23254776, 0.23256492, 0.26699012, 0.41438752, 0.59448102, 0.49043671, 0.25764383, 0.28265227, 0.49938059, 0.20976461, 0.45252510, 0.40100550, 0.24138738, 0.51419442, 0.30912931, 0.19270926, 0.46076875, 0.30703976, 0.47925088, 0.24845422, 0.37799021, 0.47495314, 0.36871868, 0.36715992, 0.48703219, 0.37048658, 0.13218894, 0.47467607, 0.33963194, 0.37156872, 0.34689429, 0.34391573, 0.40845588, 0.32799294, 0.57346577, 0.10573375, 0.45642240, 0.37851526, 0.22431294, 0.23692439, 0.33849143, 0.27381426, 0.39965198, 0.41201772, 0.33977959, 0.41796550, 0.39697713, 0.52148446, 0.18902439, 0.34751249, 0.42338701, 0.32613592, 0.50108900, 0.32613300, 0.36330014, 0.39497015, 0.33845313, 0.28011569, 0.43751187, 0.41213643, 0.24179264, 0.29566622, 0.43867378, 0.36155961, 0.49627929, 0.45379851, 0.24529567, 0.42839232, 0.12603140, 0.51068248, 0.39044941, 0.22842060, 0.28604811, 0.37920624, 0.39488357, 0.49142644, 0.19303242, 0.30509616, 0.35419436, 0.57922498, 0.34204373, 0.39049205, 0.27700611, 0.46974557, 0.37977411, 0.30986615, 0.45385451, 0.34549566, 0.44764169, 0.30396429, 0.26560635, 0.46688682, 0.34718550, 0.32563316, 0.48156394, 0.37788893, 0.45803404, 0.33401036, 0.49210787, 0.36889110, 0.37402951, 0.40458266, 0.34404844, 0.39699806, 0.42347582, 0.39383998, 0.35038586, 0.41919659, 0.47936964, 0.40647672, 0.49388580, 0.49383058, 0.43054761, 0.24525336, 0.29581170, 0.37046297, 0.49712392, 0.32128505, 0.23845768, 0.48502925, 0.33209492, 0.25236158, 0.33548368, 0.31222920, 0.11141084, 0.32674611, 0.26787342, 0.27662515, 0.48644238, 0.21554285, 0.35735214, 0.47788589, 0.39048312, 0.18714603, 0.32364833, 0.39804068, 0.46353421, 0.42251671, 0.34956771, 0.19073000, 0.37400850, 0.21161721, 0.53906612, 0.49232736, 0.48419373, 0.39627618, 0.22227510, 0.22329599, 0.38301612, 0.43001814, 0.40520558, 0.36975743, 0.32526198, 0.33220777, 0.27990257, 0.44065939, 0.45800741, 0.30765543, 0.35109396, 0.24058013, 0.15791243, 0.14863150, 0.38827008, 0.42837730, 0.46465828, 0.54745820, 0.34374755, 0.46476362, 0.44240261, 0.29813557, 0.36687965, 0.40317123, 0.40857152, 0.37164974, 0.41226479, 0.54474676, 0.19999817, 0.34599613, 0.15690204, 0.37319497, 0.49078806, 0.26886547, 0.44214884, 0.39244426, 0.46218008, 0.48172675, 0.57592097, 0.36458435, 0.41422416, 0.35969788, 0.41980495, 0.37529218, 0.30871578, 0.41929812, 0.43366044, 0.48366402, 0.32165728, 0.18604071, 0.41272197, 0.49723981, 0.33804899, 0.17234326, 0.27562049, 0.43505835, 0.21643307, 0.21458139, 0.25736381, 0.28476760, 0.31527156, 0.36744670, 0.37825692, 0.50709172, 0.43219751, 0.18977635, 0.26327341, 0.17310781, 0.38639482, 0.47189400, 0.41601084, 0.20671635, 0.56230153, 0.41836663, 0.39796858, 0.32829044, 0.30277060, 0.44775291, 0.32816917, 0.27935160, 0.35015803, 0.58615249, 0.34562539, 0.29355900, 0.28298933, 0.50509005, 0.34783965, 0.38829562, 0.37445260, 0.36001475, 0.34546856, 0.45336849, 0.33916374, 0.34239627, 0.32355265, 0.41653694, 0.47690166, 0.41311561, 0.43205490, 0.23447482, 0.41054831, 0.46526172, 0.30751118, 0.36952833, 0.26967795, 0.42790667, 0.44590686, 0.38504515, 0.36849803, 0.45946597, 0.34570813, 0.37983734, 0.39496048, 0.502255777, 0.478909263, 0.323189120, 0.374278793, 0.270686466, 0.277117398, 0.417217659, 0.426015414, 0.456790541, 0.322189137, 0.544548505, 0.253578860, 0.322975900, 0.362919840, 0.307711447, 0.364220168, 0.359471201, 0.373375671, 0.243844850, 0.290917892, 0.597136684, 0.254036144, 0.234842559, 0.455002949, 0.415709324, 0.443531265, 0.299394678, 0.273685693, 0.339751460, 0.290479625, 0.418674294, 0.367702267, 0.361132947, 0.447685742, 0.498754389, 0.270269135, 0.295705938, 0.484246007, 0.311986189, 0.304280505, 0.406538045, 0.370420053, 0.456776894, 0.457377674, 0.296940275, 0.377566228, 0.367587126, 0.427966186, 0.324805655, 0.420154277, 0.385518258, 0.434303736, 0.484742044, 0.420268053, 0.543945352, 0.614640782, 0.181089220, 0.538398740, 0.172360743, 0.401025849, 0.399125170, 0.301969996, 0.458273275, 0.352292564, 0.428755908, 0.281922745, 0.166868747, 0.383126737, 0.461700376, 0.321581217, 0.294979582, 0.645734516, 0.480530595, 0.369117806, 0.447472411, 0.367004384, 0.325281361, 0.372054625, 0.559420414, 0.420927754, 0.441013621, 0.268351322, 0.471931821, 0.377414999, 0.398822036, 0.586611520, 0.398019082, 0.332722226, 0.404158484, 0.335375695, 0.613275439, 0.442884130, 0.153235309, 0.331616205, 0.391264091, 0.357352172, 0.413816771, 0.408433901, 0.419759530, 0.228831898, 0.242195246, 0.325805978, 0.177510323, 0.287290644, 0.423608328, 0.346624498, 0.398904327, 0.312968780, 0.385049936, 0.355245850, 0.353788611, 0.622340341, 0.347298462, 0.487055228, 0.347159344, 0.642501454, 0.286065916, 0.272343803, 0.341502439, 0.502706623, 0.423043196, 0.392224835, 0.305042008, 0.413321735, 0.283283665, 0.335373677, 0.411467352, 0.016942860, 0.370115682, 0.556011768, 0.392865030, 0.301642465, 0.284051455, 0.493041813, 0.422299781, 0.283624328, 0.323970631, 0.360579274, 0.648236531, 0.492726100, 0.341422713, 0.506488260, 0.536233398, 0.399949932, 0.427973590, 0.427325204, 0.309583227, 0.320272411, 0.344996672, 0.406182681, 0.467704420, 0.443549638, 0.246331343, 0.255539278, 0.321713488, 0.241197032, 0.356909474, 0.403565810, 0.424665343, 0.449729106, 0.507606334, 0.463193632, 0.333415140, 0.447615303, 0.384183785, 0.414546057, 0.400747334, 0.537751757, 0.475605296, 0.238143678, 0.536313754, 0.421732910, 0.218056684, 0.202829815, 0.283549436, 0.275632531, 0.424492510, 0.404217234, 0.201301262, 0.424716609, 0.281339577, 0.514103586, 0.355023207, 0.140366294, 0.444906239, 0.367123147, 0.310100700, 0.457209742, 0.432594605, 0.583239047, 0.439187967, 0.411749513, 0.276365926, 0.282221109, 0.440751034, 0.607766309, 0.490552742, 0.170092443, 0.514622764, 0.387188302, 0.388399891, 0.500914171, 0.496650245, 0.589290271, 0.736033060, 0.397034165, 0.317624864, 0.298077378, 0.457904806, 0.183291084, 0.167276433, 0.458485644, 0.073808155, 0.344174108, 0.516475697, 0.422334218, 0.165966943, 0.457473721, 0.250581393, 0.481856213, 0.451418497, 0.407934704, 0.448765138, 0.343626739, 0.319897801, 0.387265723, 0.381855560, 0.314778127, 0.353790288, 0.535941667, 0.341841593, 0.407951513, 0.442824003, 0.262223904, 0.326596401, 0.321151280, 0.433902619, 0.465935076, 0.442801441, 0.310250453, 0.356294623, 0.240147508, 0.506087974, 0.367953065, 0.308278430, 0.352030300, 0.226553727, 0.398012138, 0.339787484, 0.320898727) # to run the full simulation uncomment the following line to fit the model for every dataset and not just for the first dataset #for (i_rep in seq_len(n_rep)) { for (i_rep in seq_len(1)) { df = data[[i_rep]] obj_dml_data = double_ml_data_from_data_frame(df, y_col = "y", d_cols = "d") obj_dml_plr_nonorth = DoubleMLPLR$new(obj_dml_data,
ml_g, ml_m,
n_folds=2,
score=non_orth_score,
apply_cross_fitting=FALSE)
obj_dml_plr_nonorth$fit() this_theta = obj_dml_plr_nonorth$coef
print(abs(theta_nonorth[i_rep] - this_theta))
theta_nonorth[i_rep] = this_theta
}

g_nonorth = ggplot(data.frame(theta_nonorth), aes(x = theta_nonorth)) +
geom_density(fill = "dark orange", alpha = 0.3, color = "dark orange") +
geom_vline(aes(xintercept = alpha), col = "black") +
xlim(c(0.08, 0.75)) + xlab("") + ylab("") + theme_minimal()
g_nonorth

           d
0.0008187973


The regularization bias in the simple ML-approach is caused by the slow convergence of $$\hat{\theta}$$

$|\sqrt{n} (\hat{\theta} - \theta) | \rightarrow_{P} \infty$

i.e., slower than $$1/\sqrt{n}$$. The driving factor is the bias that arises by learning $$g$$ with a random forest or any other ML technique. A heuristic illustration is given by

$\sqrt{n}(\hat{\theta} - \theta) = \underbrace{\left(\frac{1}{n} \sum_{i\in I} D_i^2\right)^{-1} \frac{1}{n} \sum_{i\in I} D_i U_i}_{=:a} + \underbrace{\left(\frac{1}{n} \sum_{i\in I} D_i^2\right)^{-1} \frac{1}{n} \sum_{i\in I} D_i (g_0(X_i) - \hat{g}_0(X_i))}_{=:b}.$

$$a$$ is approximately Gaussian under mild conditions. However, $$b$$ (the regularization bias) diverges in general.

## 1.4. Overcoming regularization bias by orthogonalization¶

To overcome the regularization bias we can partial out the effect of $$X$$ from $$D$$ to obtain the orthogonalized regressor $$V = D - m(X)$$. We then use the final estimate

$\check{\theta} = \left(\frac{1}{n} \sum_{i\in I} \hat{V}_i D_i\right)^{-1} \frac{1}{n} \sum_{i\in I} \hat{V}_i (Y_i - \hat{g}_0(X_i)).$
In [33]: import numpy as np

In [34]: np.random.seed(2222)

# to speed up the illustration we hard-code the simulation results
In [35]: theta_orth_nosplit = np.array([0.28310852, 0.2242779 , 0.21826119, 0.27480264, 0.2632724 , 0.26145622, 0.35155547, 0.29961815, 0.26635995, 0.2916099 , 0.26905356, 0.20612577, 0.26398208, 0.24738297, 0.31720634, 0.33410743, 0.22858844, 0.27328181, 0.2883969 , 0.32534549, 0.23480727, 0.27738886, 0.24176323, 0.30223882, 0.25703061, 0.24784644, 0.31566354, 0.27268702, 0.26854096, 0.17321738, 0.25972241, 0.22442174, 0.29994908, 0.32036162, 0.32746914, 0.24166733, 0.19887599, 0.30592848, 0.22804781, 0.23233486, 0.30921309, 0.23463971, 0.29075226, 0.1809537 , 0.27576646, 0.27086624, 0.24678385, 0.20240443, 0.25876583, 0.25828577, 0.34559146, 0.26086781, 0.33256474, 0.29259962, 0.25623473, 0.26968172, 0.26378077, 0.19208867, 0.24181579, 0.29534961, 0.2733621 , 0.22095648, 0.28348618, 0.30043176, 0.25261295, 0.26165325, 0.28380731, 0.23151629, 0.25394419, 0.23464047, 0.24771   , 0.29674846, 0.20385661, 0.21034693, 0.22878356, 0.29613812, 0.29508396, 0.17701553, 0.27132781, 0.23536996, 0.26647282, 0.19601602, 0.21138042, 0.25673211, 0.2520547 , 0.24963577, 0.21469675, 0.27812185, 0.28731821, 0.21461844, 0.32501942, 0.27049369, 0.25925077, 0.15021988, 0.22308663, 0.25618922, 0.29798101, 0.27308533, 0.28463352, 0.27359688, 0.23853733, 0.25253259, 0.28388183, 0.27816227, 0.19846811, 0.20672709, 0.24345814, 0.21094738, 0.29825134, 0.27913302, 0.25599672, 0.28127914, 0.21570845, 0.23568041, 0.27717801, 0.24924811, 0.29236107, 0.22863753, 0.28686067, 0.21806731, 0.19096662, 0.1991767 , 0.29704411, 0.26143366, 0.2855753 , 0.22812379, 0.31258998, 0.24737797, 0.29765883, 0.26660461, 0.22636728, 0.19561781, 0.30000204, 0.28156593, 0.2518048 , 0.24588117, 0.29438604, 0.27202006, 0.19251048, 0.29858927, 0.30207488, 0.27311435, 0.22591445, 0.23439006, 0.22010727, 0.19731043, 0.28078517, 0.2686372 , 0.263956  , 0.29857334, 0.2474458 , 0.24640703, 0.21223515, 0.22386055, 0.2255213 , 0.15326043, 0.2518272 , 0.31929998, 0.26351882, 0.21678552, 0.29317478, 0.26989617, 0.2818495 , 0.2848359 , 0.21478301, 0.22616049, 0.21466051, 0.26933743, 0.26859455, 0.23889104, 0.29573917, 0.24851225, 0.28498028, 0.27178069, 0.21132864, 0.24681366, 0.17802672, 0.27507623, 0.30868363, 0.23293956, 0.30121311, 0.27621463, 0.26063687, 0.23322101, 0.24393135, 0.23227465, 0.22844058, 0.30839355, 0.22210363, 0.22822178, 0.24866478, 0.21586847, 0.26500871, 0.2471022 , 0.20997401, 0.23643077, 0.25231603, 0.27148353, 0.28322569, 0.33346698, 0.25167926, 0.26491853, 0.24722962, 0.27637599, 0.23881742, 0.2296211 , 0.19967581, 0.31846702, 0.29736598, 0.27647978, 0.30541239, 0.23427654, 0.24818056, 0.26726521, 0.26369045, 0.24638188, 0.27767191, 0.30294559, 0.28852899, 0.27617697, 0.21988934, 0.292125  , 0.18724467, 0.26117014, 0.24419148, 0.2871237 , 0.24905991, 0.22842965, 0.27704528, 0.25578749, 0.25142395, 0.36927582, 0.22390373, 0.21829721, 0.12744814, 0.26925737, 0.20636333, 0.20155245, 0.21186804, 0.26988056, 0.24059909, 0.31412907, 0.26759994, 0.22547844, 0.23929809, 0.26712295, 0.29086087, 0.31798801, 0.29562617, 0.33280922, 0.26751236, 0.28698399, 0.22986446, 0.2763425 , 0.19818218, 0.24978374, 0.2734057 , 0.25108121, 0.29437982, 0.20228561, 0.22372583, 0.25953038, 0.3081489 , 0.30320545, 0.21424474, 0.27065109, 0.26593162, 0.29835337, 0.24485177, 0.30415425, 0.28980662, 0.2680405 , 0.239961  , 0.24617116, 0.28658336, 0.29611991, 0.25713896, 0.30998392, 0.2665378 , 0.18528186, 0.2424845 , 0.25698249, 0.33527911, 0.3207351 , 0.23184278, 0.26313123, 0.28758601, 0.25173612, 0.27760017, 0.26816815, 0.28290835, 0.27865219, 0.35303055, 0.25650497, 0.16584378, 0.25912287, 0.18661047, 0.24201838, 0.19715925, 0.25251827, 0.23211738, 0.33786812, 0.2259777 , 0.24312618, 0.26597777, 0.20004984, 0.18565445, 0.2554961 , 0.17862975, 0.25421527, 0.19827303, 0.27165019, 0.266252  , 0.30177934, 0.29902409, 0.25955909, 0.19670987, 0.22927377, 0.24951701, 0.27330244, 0.22528742, 0.26322332, 0.30611685, 0.24612744, 0.27514948, 0.28442492, 0.27129947, 0.29369819, 0.2795243 , 0.29063484, 0.24778475, 0.29604971, 0.22014627, 0.27125729, 0.22028222, 0.23937117, 0.35621441, 0.22856674, 0.21706003, 0.27878548, 0.23050219, 0.27815778, 0.29854744, 0.32191927, 0.22175815, 0.22458026, 0.25637069, 0.19114296, 0.29236829, 0.25682208, 0.29432575, 0.23749481, 0.28604151, 0.28598776, 0.29619051, 0.25680905, 0.34011962, 0.24549912, 0.29141369, 0.23814109, 0.21578371, 0.2807999 , 0.27404827, 0.28924312, 0.20979212, 0.27621514, 0.32499504, 0.28444786, 0.20769076, 0.21044053, 0.21242127, 0.2025243 , 0.26759374, 0.29333089, 0.31110005, 0.30975388, 0.25719257, 0.28956785, 0.2484451 , 0.25275622, 0.29471806, 0.23193147, 0.25748414, 0.16584411, 0.31336108, 0.22796508, 0.32666538, 0.22967239, 0.2605007 , 0.21030846, 0.17930794, 0.25799981, 0.22374191, 0.24851312, 0.18130437, 0.26671058, 0.28468216, 0.26699029, 0.23856576, 0.2378861 , 0.29746227, 0.22210649, 0.25561485, 0.30647856, 0.31982858, 0.23269608, 0.23954949, 0.27028194, 0.2659938 , 0.16761426, 0.23152631, 0.26632464, 0.25441972, 0.31960062, 0.2765675 , 0.26511514, 0.22645075, 0.29570968, 0.222362  , 0.25907818, 0.19219834, 0.18055943, 0.27491122, 0.26250238, 0.32790241, 0.24976504, 0.22750637, 0.33063588, 0.29926466, 0.2392139 , 0.27519006, 0.29970428, 0.25545158, 0.29449147, 0.23462131, 0.24624584, 0.25795428, 0.25209296, 0.24903425, 0.2664015 , 0.24116091, 0.23901285, 0.28603948, 0.26749332, 0.24186271, 0.25248758, 0.19619871, 0.3029078 , 0.2967233 , 0.3082407 , 0.21993216, 0.2433419 , 0.288618  , 0.21555319, 0.25470933, 0.23773486, 0.29546983, 0.21327642, 0.24956888, 0.2889786 , 0.24442302, 0.33887585, 0.31841493, 0.24695447, 0.32795625, 0.20303785, 0.27606344, 0.26365563, 0.31353331, 0.22108299, 0.34605167, 0.18610914, 0.28660735, 0.32646627, 0.298221  , 0.27317117, 0.3345105 , 0.25627246, 0.32404571, 0.30980458, 0.26954175, 0.20478813, 0.3305093 , 0.29632759, 0.2263084 , 0.26588734, 0.32757012, 0.20919784, 0.22280691, 0.24741845, 0.23078147, 0.23477978, 0.2658129 , 0.24613582, 0.25510153, 0.24454531, 0.32516082, 0.30274893, 0.23969294, 0.28680135, 0.24591505, 0.21699139, 0.30187635, 0.29572537, 0.2630132 , 0.26787657, 0.24316903, 0.26117894, 0.299745  , 0.30256152, 0.31785968, 0.29531903, 0.3064115 , 0.29840839, 0.20857716, 0.27004916, 0.25333668, 0.20998201, 0.26692238, 0.2723465 , 0.32207802, 0.22611339, 0.20679392, 0.24201841, 0.26512521, 0.25212648, 0.2576815 , 0.23881843, 0.23155785, 0.36747982, 0.19636713, 0.32124027, 0.2427754 , 0.2011566 , 0.33390234, 0.18539832, 0.30159911, 0.20772415, 0.2356763 , 0.34007911, 0.29580993, 0.22988797, 0.21634261, 0.26545917, 0.16314891, 0.33988473, 0.29242298, 0.22733269, 0.24611494, 0.30183313, 0.32251464, 0.26732256, 0.24792092, 0.2772301 , 0.28939186, 0.26631433, 0.28354286, 0.33360631, 0.24205665, 0.1961794 , 0.24519602, 0.23249065, 0.23737376, 0.25656539, 0.21403472, 0.29415284, 0.26959061, 0.23195155, 0.26536177, 0.24566337, 0.23165363, 0.22660996, 0.26522449, 0.33309614, 0.21471191, 0.30851745, 0.25249242, 0.31809431, 0.22793364, 0.27710466, 0.27294355, 0.29596034, 0.29298504, 0.2859605 , 0.27495038, 0.26136098, 0.2934716 , 0.27540508, 0.2690389 , 0.31972379, 0.22868217, 0.29370394, 0.20299354, 0.34906811, 0.28639738, 0.37465074, 0.2956081 , 0.2842431 , 0.24320374, 0.27014277, 0.2242631 , 0.28506609, 0.25101436, 0.24047286, 0.25575328, 0.22997231, 0.2205248 , 0.21703479, 0.20778374, 0.25258876, 0.2401245 , 0.29990474, 0.26323848, 0.25627857, 0.26284862, 0.2441438 , 0.23134306, 0.22144132, 0.30448487, 0.25150713, 0.22797692, 0.24349524, 0.34462206, 0.27105137, 0.2892143 , 0.26127535, 0.25287297, 0.23561471, 0.14121307, 0.30581663, 0.29274273, 0.24767589, 0.33147642, 0.27992392, 0.33720554, 0.2090374 , 0.27164353, 0.31701844, 0.28533359, 0.25799022, 0.3070244 , 0.17683673, 0.21358908, 0.26084117, 0.22207999, 0.3236883 , 0.33255362, 0.24625949, 0.24396244, 0.30353548, 0.32189553, 0.26497649, 0.23178877, 0.24712478, 0.29951475, 0.24184459, 0.24421118, 0.30500367, 0.25970607, 0.27909354, 0.25431678, 0.27946877, 0.23949309, 0.26691102, 0.28535357, 0.26128905, 0.25824165, 0.28183132, 0.17514963, 0.33431912, 0.18331401, 0.26675644, 0.30160769, 0.2497697 , 0.22613143, 0.26237311, 0.26848567, 0.28228926, 0.29263579, 0.29236556, 0.1840749 , 0.24696389, 0.26452469, 0.28719122, 0.23128346, 0.24169805, 0.23304394, 0.218957  , 0.25475361, 0.35289907, 0.2867506 , 0.26357185, 0.2270227 , 0.25602031, 0.27716229, 0.20449755, 0.26554645, 0.25110001, 0.24079137, 0.30162768, 0.28170507, 0.17551756, 0.32547549, 0.22677441, 0.14042389, 0.17337534, 0.26256224, 0.23707877, 0.22717987, 0.24024014, 0.25177129, 0.27503805, 0.28114949, 0.25896209, 0.2662017 , 0.21796715, 0.23287975, 0.20286762, 0.23630051, 0.24797168, 0.23176171, 0.3070345 , 0.24794182, 0.21011151, 0.23875281, 0.25753579, 0.21149103, 0.23827549, 0.21968091, 0.26969627, 0.26792777, 0.20638949, 0.27621866, 0.2604688 , 0.26865353, 0.2157272 , 0.15645571, 0.26005388, 0.24269695, 0.26726758, 0.29685328, 0.25025327, 0.25305107, 0.22538429, 0.23264148, 0.26118271, 0.22936266, 0.24315368, 0.26195489, 0.28365961, 0.27980084, 0.26609696, 0.1696261 , 0.33175361, 0.28457782, 0.29927934, 0.23976468, 0.23491615, 0.18087741, 0.20349075, 0.21891415, 0.2522452 , 0.27317506, 0.20354548, 0.29480749, 0.23066701, 0.24240663, 0.28959264, 0.23725127, 0.3009052 , 0.3272782 , 0.33884528, 0.32950748, 0.26603133, 0.28544335, 0.31174843, 0.35525527, 0.16409966, 0.20273623, 0.21960992, 0.20445596, 0.25733506, 0.26104803, 0.25997223, 0.2684171 , 0.27375778, 0.19854972, 0.28553536, 0.25885901, 0.28304118, 0.2621129 , 0.27795949, 0.28436075, 0.25445093, 0.24717689, 0.22179333, 0.31690662, 0.21867459, 0.27198495, 0.25701943, 0.28360551, 0.23875228, 0.22175629, 0.21333009, 0.2300189 , 0.25209904, 0.28502759, 0.27892907, 0.27256734, 0.23018934, 0.25019992, 0.24872246, 0.26869467, 0.22006607, 0.31613469, 0.19378017, 0.25437549, 0.31010428, 0.2239705 , 0.20986456, 0.2981559 , 0.22735091, 0.2385243 , 0.28349625, 0.26244455, 0.22222827, 0.25961584, 0.23382021, 0.28271075, 0.27364524, 0.22208577, 0.22338154, 0.25133704, 0.2927259 , 0.27249243, 0.19810369, 0.29739084, 0.23977764, 0.33163001, 0.20838183, 0.32524989, 0.22693308, 0.24781074, 0.26384576, 0.24114416, 0.22417866, 0.26879515, 0.2671183 , 0.27647383, 0.21137266, 0.25714512, 0.31201653, 0.28097195, 0.23299879, 0.2663719 , 0.28909349, 0.2835265 , 0.25468493, 0.24928326, 0.26264776, 0.28256157, 0.19424368, 0.2629451 , 0.29564766, 0.23307118, 0.21502833, 0.28635937, 0.28488341, 0.2700886 , 0.28704458, 0.27922478, 0.17263791, 0.28635712, 0.23075424, 0.17741433, 0.23449618, 0.18159064, 0.25603823, 0.24429592, 0.252812  , 0.27307543, 0.34339119, 0.252644  , 0.29226866, 0.19520718, 0.27469428, 0.28431326, 0.2082317 , 0.28421696, 0.3070267 , 0.30473623, 0.27110717, 0.28557713, 0.25306611, 0.20115915, 0.28109259, 0.23066984, 0.29306542, 0.23246166, 0.26619353, 0.26780462, 0.21972068, 0.21848152, 0.33710686, 0.2530146 , 0.25794811, 0.26789696, 0.20139948, 0.31620582, 0.18905824, 0.30757573, 0.25823602, 0.32225168, 0.29416294, 0.33819316, 0.33667058, 0.24208325, 0.24359405, 0.26374286, 0.28239228, 0.23335644, 0.27874058, 0.30653043, 0.26781161, 0.2588007 , 0.26839054, 0.28338128, 0.28230784, 0.28294972, 0.26593038, 0.33051548, 0.31535588, 0.22534891, 0.29542099, 0.25698647, 0.30864245, 0.21753042, 0.25400367, 0.22775759, 0.19603169, 0.23960378, 0.31154354, 0.27754087, 0.20881702, 0.28891996, 0.28937791, 0.29420089, 0.26498318, 0.30885406, 0.26559376, 0.27107953, 0.31337713, 0.21710285, 0.24758666, 0.24602542, 0.2719711 , 0.29087773, 0.22520773, 0.22748796, 0.20393595, 0.28756757, 0.14159469, 0.30246539, 0.20772598, 0.2444757 , 0.25797414, 0.27913936, 0.13832664, 0.15671679, 0.3150215 , 0.21286612, 0.30024509, 0.25108497, 0.33896346, 0.23340797, 0.32430489, 0.27986208, 0.24811693, 0.18176737, 0.23715368, 0.19188467, 0.28837014, 0.28031912, 0.26374323, 0.2776042 , 0.19989007, 0.27678714, 0.29959389, 0.19771432, 0.2368231 , 0.36130171, 0.23823292, 0.22617482, 0.24363743, 0.28811019, 0.31579602, 0.25930383, 0.22536299, 0.29135006, 0.28875076])

# to run the full simulation uncomment the following line to fit the model for every dataset and not just for the first dataset
#for i_rep in range(n_rep):
In [36]: for i_rep in range(1):
....:     (x, y, d) = data[i_rep]
....:     obj_dml_data = DoubleMLData.from_arrays(x, y, d)
....:     obj_dml_plr_orth_nosplit = DoubleMLPLR(obj_dml_data,
....:                                            ml_g, ml_m,
....:                                            n_folds=1,
....:                                            score='IV-type',
....:                                            apply_cross_fitting=False)
....:     obj_dml_plr_orth_nosplit.fit()
....:     this_theta = obj_dml_plr_orth_nosplit.coef[0]
....:     print(np.abs(theta_orth_nosplit[i_rep] - this_theta))
....:     theta_orth_nosplit[i_rep] = this_theta
....:
0.0004266738001413861

In [37]: ax = sns.kdeplot(theta_orth_nosplit, shade=True, color=colors[2])

In [38]: ax.axvline(0.5, color='k', label='True theta');

library(data.table)
lgr::get_logger("mlr3")$set_threshold("warn") set.seed(2222) # to speed up the illustration we hard-code the simulation results theta_orth_nosplit = c(0.24404981, 0.24200765, 0.27908831, 0.25502043, 0.32824182, 0.26923132, 0.24104406, 0.20969115, 0.32043631, 0.19615339, 0.26752732, 0.28054535, 0.20565819, 0.23094579, 0.30052261, 0.22506210, 0.30899243, 0.33166896, 0.29901923, 0.21645924, 0.19363222, 0.25194347, 0.20642525, 0.24555231, 0.22453258, 0.30131307, 0.23932457, 0.24134974, 0.32292519, 0.24820202, 0.29051209, 0.19341549, 0.25560635, 0.24635088, 0.25058896, 0.19513198, 0.23325175, 0.25732967, 0.25630722, 0.23618513, 0.17672957, 0.21211532, 0.26970909, 0.23979100, 0.24278216, 0.31524081, 0.21496712, 0.23975950, 0.22430487, 0.25032633, 0.31171158, 0.28673753, 0.23617691, 0.30279132, 0.25171843, 0.21937215, 0.24240521, 0.23905356, 0.18858651, 0.20627836, 0.18322578, 0.25066665, 0.27527609, 0.27380098, 0.16270080, 0.22131889, 0.22989677, 0.25631445, 0.30572004, 0.28186739, 0.28453591, 0.31102750, 0.28063015, 0.25876131, 0.29582343, 0.24814810, 0.35109199, 0.25108377, 0.22487320, 0.30794975, 0.23894645, 0.26638033, 0.23674057, 0.20538273, 0.19608607, 0.33079935, 0.19395816, 0.21966814, 0.18736528, 0.21029872, 0.29902640, 0.26053716, 0.25603558, 0.21075900, 0.20782755, 0.28236968, 0.25193825, 0.28224143, 0.23540541, 0.26424208, 0.29290447, 0.24258066, 0.27822969, 0.20114274, 0.23819599, 0.25825846, 0.30886038, 0.21687679, 0.26809187, 0.29897851, 0.28606557, 0.25426056, 0.17297763, 0.22302356, 0.24706098, 0.24361315, 0.20613373, 0.20395747, 0.26581544, 0.27803372, 0.28981682, 0.28375231, 0.26902671, 0.22007559, 0.21972414, 0.22691012, 0.23230200, 0.19660759, 0.22971657, 0.24840646, 0.27392351, 0.23431585, 0.23109943, 0.22423472, 0.17548714, 0.27187182, 0.23154822, 0.23570134, 0.20276194, 0.20357300, 0.29668416, 0.23894281, 0.29103412, 0.26916173, 0.20825452, 0.27174241, 0.28702397, 0.26323370, 0.20993333, 0.22861364, 0.28225493, 0.25282767, 0.25466747, 0.20584595, 0.25209117, 0.26527392, 0.24857230, 0.30070528, 0.29237504, 0.28291041, 0.26424070, 0.20749123, 0.33763761, 0.17805976, 0.23412056, 0.31256005, 0.24081702, 0.24795085, 0.29925883, 0.25657520, 0.17849887, 0.27372626, 0.23904392, 0.22488373, 0.24942245, 0.29803576, 0.23674057, 0.27779304, 0.25638963, 0.22914358, 0.22338097, 0.23024613, 0.22301133, 0.29410433, 0.30106813, 0.27284832, 0.24042849, 0.24791942, 0.25071209, 0.25156142, 0.23516980, 0.24695011, 0.25794186, 0.27033305, 0.27898959, 0.22825721, 0.26305627, 0.28653743, 0.24861610, 0.29348304, 0.22902869, 0.30333429, 0.26935249, 0.19811759, 0.27767260, 0.19347321, 0.33130437, 0.25953422, 0.26075776, 0.17054191, 0.22197221, 0.26321300, 0.27792267, 0.20024470, 0.31713220, 0.23528659, 0.28108999, 0.21161821, 0.18928942, 0.25071159, 0.32293010, 0.18651541, 0.20749911, 0.22686192, 0.30217232, 0.27595429, 0.26376699, 0.28513188, 0.29005017, 0.22022413, 0.18666795, 0.18394129, 0.22996187, 0.29361925, 0.24450357, 0.29357722, 0.22286200, 0.28265768, 0.22675825, 0.27649261, 0.28951131, 0.28271866, 0.17806041, 0.21163621, 0.30509409, 0.26342174, 0.32677576, 0.24797358, 0.23200651, 0.27904545, 0.26504825, 0.19402431, 0.26306555, 0.31078689, 0.24485790, 0.26520010, 0.21393591, 0.25235737, 0.23623305, 0.21269470, 0.36499369, 0.21848834, 0.28736613, 0.25397938, 0.25896726, 0.14757031, 0.15458744, 0.25966165, 0.29306232, 0.29959702, 0.34040452, 0.24615155, 0.16853516, 0.28525494, 0.23482806, 0.23951146, 0.26795442, 0.26138905, 0.28521849, 0.30644821, 0.22308750, 0.20662353, 0.26295613, 0.23490771, 0.20767176, 0.25672486, 0.23762899, 0.24988185, 0.21882009, 0.32730028, 0.31310884, 0.25821877, 0.23069627, 0.22345636, 0.19806570, 0.29756479, 0.21450103, 0.27358563, 0.24101731, 0.23376150, 0.22875492, 0.30380284, 0.27886358, 0.25644140, 0.27983788, 0.29336975, 0.20603135, 0.23124943, 0.25156872, 0.30053623, 0.27771452, 0.19847193, 0.25991230, 0.25088226, 0.25607595, 0.29055568, 0.20725805, 0.20844012, 0.26323932, 0.26251734, 0.29311885, 0.29691861, 0.28589456, 0.28678294, 0.23915972, 0.26246128, 0.24366650, 0.25425399, 0.31767331, 0.25261640, 0.21726079, 0.30237963, 0.22177598, 0.19588260, 0.23795208, 0.26311371, 0.28360425, 0.21153313, 0.17872878, 0.23587613, 0.25150900, 0.25867927, 0.26743636, 0.19428422, 0.24070189, 0.26749686, 0.21719999, 0.26282296, 0.23008293, 0.28084342, 0.22708225, 0.26011481, 0.25512443, 0.24929613, 0.26494690, 0.32442653, 0.24635223, 0.23136118, 0.29793829, 0.22573153, 0.20318348, 0.26693828, 0.24952603, 0.27215040, 0.26322105, 0.22329061, 0.27891631, 0.27127217, 0.27125980, 0.27248351, 0.22775018, 0.25404143, 0.25163026, 0.30982078, 0.20748986, 0.27073473, 0.23114844, 0.20224835, 0.30671944, 0.19242940, 0.24885178, 0.22356225, 0.29483575, 0.25740248, 0.18617692, 0.24523393, 0.25576838, 0.35176784, 0.25000509, 0.29496086, 0.18229425, 0.31624955, 0.21398406, 0.31201873, 0.27873516, 0.29288879, 0.21909350, 0.33838391, 0.28401419, 0.28814919, 0.25356693, 0.28313276, 0.29096662, 0.24443356, 0.27196073, 0.22707857, 0.23169359, 0.23623725, 0.29841002, 0.19643420, 0.22633211, 0.28084501, 0.26918329, 0.36923637, 0.21106338, 0.27083820, 0.23144420, 0.31389920, 0.22205859, 0.22125093, 0.26983642, 0.26984812, 0.27143384, 0.22440664, 0.24382201, 0.29489124, 0.27572998, 0.23011901, 0.21742103, 0.24521077, 0.24721123, 0.29630124, 0.29493903, 0.28724526, 0.27999346, 0.24368907, 0.22984274, 0.17392723, 0.27441598, 0.25581560, 0.22590773, 0.28188015, 0.22709537, 0.31067150, 0.32512125, 0.23057776, 0.30177850, 0.23242622, 0.28326166, 0.29190056, 0.25107901, 0.23130449, 0.24929647, 0.19796943, 0.26118980, 0.28361317, 0.28984450, 0.24239718, 0.23415392, 0.27891457, 0.31708859, 0.25101173, 0.24828103, 0.26219254, 0.26629355, 0.30090049, 0.21231956, 0.28541164, 0.31735485, 0.23788811, 0.21526438, 0.22865108, 0.26906438, 0.21838840, 0.26907421, 0.26570784, 0.20624892, 0.30029962, 0.34755127, 0.31459386, 0.28203083, 0.26903597, 0.29458419, 0.23009954, 0.24293185, 0.29547741, 0.22776509, 0.25928743, 0.27324063, 0.19938081, 0.34164972, 0.27462158, 0.22935702, 0.32378804, 0.25564062, 0.23829308, 0.17257210, 0.27254683, 0.20348844, 0.31883511, 0.30223972, 0.28047451, 0.25603642, 0.27022816, 0.28062654, 0.25154276, 0.25218666, 0.26535623, 0.21176103, 0.21301494, 0.26220526, 0.27431630, 0.27729789, 0.24779401, 0.24868582, 0.18360005, 0.26786428, 0.24958852, 0.22578138, 0.27438830, 0.32905933, 0.31888969, 0.24873045, 0.28213371, 0.20563072, 0.24977431, 0.30339369, 0.24085435, 0.17003776, 0.25333376, 0.28797501, 0.30977813, 0.25259406, 0.30296617, 0.26677033, 0.22128846, 0.30456640, 0.24690890, 0.30914977, 0.33190582, 0.26943920, 0.32514576, 0.25079613, 0.28332654, 0.22076585, 0.27260810, 0.21275210, 0.32582630, 0.28814724, 0.22820021, 0.23679701, 0.26929291, 0.20674902, 0.21578408, 0.23102393, 0.19422745, 0.26693484, 0.24910152, 0.25459030, 0.28373457, 0.23307690, 0.22377395, 0.22574906, 0.29702891, 0.25427968, 0.21206913, 0.24253494, 0.23148299, 0.28956044, 0.27245119, 0.27969498, 0.21298051, 0.27298282, 0.26791011, 0.23978439, 0.26989927, 0.20125381, 0.26407324, 0.30557669, 0.22319726, 0.31493078, 0.22910660, 0.21667511, 0.23672426, 0.25941675, 0.29885728, 0.21517201, 0.28029613, 0.32394264, 0.27670934, 0.34767623, 0.25234206, 0.28996521, 0.27791141, 0.25572864, 0.24964783, 0.21480398, 0.28467806, 0.28029405, 0.19916957, 0.30238731, 0.25757664, 0.21437019, 0.22555528, 0.28787462, 0.30702164, 0.23218925, 0.23998290, 0.24108701, 0.19807579, 0.27532625, 0.29797673, 0.33230228, 0.24863314, 0.26096881, 0.31315212, 0.24171227, 0.25872465, 0.22154767, 0.27599649, 0.26455210, 0.23163402, 0.27558069, 0.27989725, 0.27851760, 0.21170799, 0.28618903, 0.19861032, 0.22324798, 0.30309738, 0.17963283, 0.26532334, 0.24133140, 0.26011628, 0.31859660, 0.22378082, 0.26713240, 0.27861435, 0.21975233, 0.32211484, 0.28379428, 0.24634791, 0.21500082, 0.24459111, 0.22421237, 0.25253544, 0.23511794, 0.30314989, 0.17381773, 0.25880032, 0.21234982, 0.23865437, 0.25171010, 0.30581337, 0.28307510, 0.28132029, 0.25098349, 0.24429769, 0.32528412, 0.24508261, 0.27625946, 0.26367128, 0.30872965, 0.24758068, 0.31172295, 0.24418416, 0.20467152, 0.23338737, 0.24866117, 0.27964005, 0.27014072, 0.30935446, 0.25824221, 0.27114065, 0.24802035, 0.22854418, 0.22945380, 0.21464254, 0.25982662, 0.30725683, 0.22635262, 0.24299934, 0.24783000, 0.31402718, 0.23889412, 0.29329641, 0.24323276, 0.25997723, 0.29938332, 0.21108355, 0.17714399, 0.23398678, 0.31824499, 0.22832691, 0.23374580, 0.23741578, 0.19794363, 0.19985305, 0.25941090, 0.33144872, 0.27671553, 0.23529043, 0.25156616, 0.21581062, 0.14069085, 0.28701229, 0.28376436, 0.25324103, 0.24521140, 0.26348733, 0.31471963, 0.27928279, 0.23140739, 0.24005349, 0.27719299, 0.22549446, 0.17181816, 0.26914335, 0.26147563, 0.23488390, 0.20516042, 0.23104556, 0.33939887, 0.22747521, 0.26699641, 0.26940843, 0.24175948, 0.27359773, 0.26232683, 0.27280967, 0.28311917, 0.24809043, 0.23792607, 0.29930390, 0.28970429, 0.20569726, 0.25444149, 0.33283531, 0.32671300, 0.25556005, 0.21977557, 0.21200236, 0.24466482, 0.21267102, 0.28793926, 0.21491347, 0.27112680, 0.27976902, 0.28284870, 0.25920649, 0.27500549, 0.28002343, 0.30423065, 0.32924577, 0.17173157, 0.18553040, 0.23845241, 0.22745331, 0.32056756, 0.22080223, 0.29517097, 0.30737837, 0.25486917, 0.23718771, 0.28089043, 0.31238705, 0.25876756, 0.24292196, 0.22375197, 0.24283881, 0.35410498, 0.23368806, 0.19522686, 0.25769016, 0.25454050, 0.31319123, 0.17947042, 0.19190434, 0.30265186, 0.19946304, 0.30086939, 0.20109633, 0.25334314, 0.20911893, 0.28693220, 0.22298880, 0.27593279, 0.30214048, 0.26325374, 0.21294424, 0.27791242, 0.26954320, 0.29819193, 0.17927875, 0.23121506, 0.26231314, 0.25083801, 0.28033401, 0.26820107, 0.27022253, 0.22792994, 0.21512806, 0.30392373, 0.37016435, 0.31230905, 0.29721586, 0.27724949, 0.25459834, 0.15951586, 0.19782845, 0.21939142, 0.23747047, 0.24440075, 0.28495355, 0.29067143, 0.22289957, 0.19727838, 0.26182810, 0.26700056, 0.20850858, 0.22248073, 0.30294435, 0.28396686, 0.25325660, 0.28110754, 0.30402911, 0.25165942, 0.28631568, 0.21262876, 0.26504292, 0.23227707, 0.20243507, 0.24853321, 0.19272924, 0.26087873, 0.32408585, 0.21510694, 0.27491827, 0.25804386, 0.22382809, 0.30728360, 0.27347877, 0.24220788, 0.19617423, 0.24985472, 0.29069347, 0.32784967, 0.23223983, 0.28298148, 0.24628940, 0.30607884, 0.23767203, 0.20429131, 0.26093432, 0.22608648, 0.22543390, 0.23870808, 0.22717988, 0.30429532, 0.21939188, 0.24270492, 0.32245359, 0.30286073, 0.22104531, 0.22910319, 0.32607062, 0.21411771, 0.25257003, 0.24598362, 0.23180878, 0.27113287, 0.25057170, 0.19189994, 0.23257955, 0.27667730, 0.22466231, 0.25881201, 0.23175854, 0.24960139, 0.32826814, 0.28951057, 0.21486691, 0.17778950, 0.26021780, 0.27643750, 0.17872446, 0.28002911, 0.21805482, 0.38678121, 0.25471786, 0.21297770, 0.32726096, 0.28189860, 0.22361958, 0.28245412, 0.26690088, 0.30962131, 0.25677141, 0.23710782, 0.23152883, 0.27632201, 0.27235424, 0.21711356, 0.21794149, 0.20043823, 0.20760472, 0.20909603, 0.31225668, 0.27529778, 0.19133858, 0.24952837, 0.32137579, 0.25944996, 0.31970595, 0.24870416, 0.26509979, 0.26355110, 0.29517943, 0.34637904, 0.24146828, 0.28212012, 0.23472525, 0.24156734, 0.23788934, 0.24383514, 0.14396294, 0.33060120, 0.29607903, 0.20853050, 0.28021850, 0.23598870, 0.25251283, 0.26717264, 0.20898575, 0.24692804, 0.28417276, 0.21140865, 0.31535353, 0.25921238, 0.27417029, 0.26054656, 0.29587580, 0.20549704, 0.27341285, 0.29240708, 0.32059263, 0.23118675, 0.17993850, 0.26216515, 0.28442858, 0.23133134, 0.26663198, 0.33022619, 0.31993129, 0.30955249, 0.33249114, 0.29334380, 0.19866766, 0.30970885, 0.21414314, 0.18668980, 0.20812096, 0.27479207, 0.25533000, 0.24358777, 0.26188038, 0.25086771, 0.25654017, 0.22080933, 0.19350181, 0.25724694, 0.26805649, 0.13882816, 0.26080592, 0.27079933, 0.24554024, 0.23415862, 0.25285817, 0.27472101, 0.24890736, 0.35753835, 0.26952921, 0.22867552, 0.21595742, 0.25398338, 0.26967079, 0.27074357, 0.21771453, 0.28066849, 0.18347077, 0.24902906, 0.23107295, 0.35380816, 0.26439250, 0.23904248, 0.26452052, 0.28343627, 0.24538294, 0.25790816, 0.21900178) # to run the full simulation uncomment the following line to fit the model for every dataset and not just for the first dataset #for (i_rep in seq_len(n_rep)){ for (i_rep in seq_len(1)) { df = data[[i_rep]] obj_dml_data = double_ml_data_from_data_frame(df, y_col = "y", d_cols = "d") obj_dml_plr_orth_nosplit = DoubleMLPLR$new(obj_dml_data,
ml_g, ml_m,
n_folds=1,
score='IV-type',
apply_cross_fitting=FALSE)
obj_dml_plr_orth_nosplit$fit() this_theta = obj_dml_plr_orth_nosplit$coef
print(abs(theta_orth_nosplit[i_rep] - this_theta))
theta_orth_nosplit[i_rep] = this_theta
}

g_nosplit = ggplot(data.frame(theta_orth_nosplit), aes(x = theta_orth_nosplit)) +
geom_density(fill = "dark green", alpha = 0.3, color = "dark green") +
geom_vline(aes(xintercept = alpha), col = "black") +
xlim(c(0.08, 0.75)) + xlab("") + ylab("") + theme_minimal()
g_nosplit

          d
3.58201e-09


If the nuisance models $$\hat{g}_0()$$ and $$\hat{m}()$$ are estimated on the whole dataset, which is also used for obtaining the final estimate $$\check{\theta}$$, another bias is observed.

## 1.5. Sample splitting to remove bias induced by overfitting¶

Using sample splitting, i.e., estimate the nuisance models $$\hat{g}_0()$$ and $$\hat{m}()$$ on one part of the data (training data) and estimate $$\check{\theta}$$ on the other part of the data (test data), overcomes the bias induced by overfitting. We can exploit the benefits of cross-fitting by switching the role of the training and test sample. Cross-fitting performs well empirically because the entire sample can be used for estimation.

In [39]: import numpy as np

In [40]: np.random.seed(3333)

# to speed up the illustration we hard-code the simulation results
In [41]: theta_dml = np.array([0.52231688, 0.43821994, 0.36741092, 0.52884091, 0.5015594 , 0.58615797, 0.50844654, 0.54049318, 0.54273396, 0.53888897, 0.507104  , 0.40035828, 0.48716359, 0.46810072, 0.64261576, 0.71117208, 0.42699728, 0.4835898 , 0.49635877, 0.50369895, 0.5347252 , 0.49725455, 0.53288677, 0.58781331, 0.58458898, 0.577733  , 0.63314775, 0.56648841, 0.51656325, 0.5853161 , 0.55397083, 0.49646073, 0.70452214, 0.57921975, 0.66033219, 0.4141339 , 0.39301364, 0.59563687, 0.48364131, 0.48707462, 0.59045756, 0.50181798, 0.55173923, 0.37780184, 0.49094078, 0.56072755, 0.51914519, 0.47253627, 0.56482564, 0.45755482, 0.58579367, 0.45175687, 0.55837968, 0.59198114, 0.54112865, 0.51840052, 0.45087786, 0.43342863, 0.51636424, 0.62517089, 0.47311126, 0.39433771, 0.62839357, 0.55921553, 0.54676839, 0.5514561 , 0.57210741, 0.45656539, 0.50132608, 0.52485313, 0.52286524, 0.50770619, 0.35230866, 0.47302836, 0.47157604, 0.57120419, 0.59794613, 0.39487438, 0.44932531, 0.42548093, 0.55208887, 0.41394061, 0.43047632, 0.52314211, 0.46423066, 0.55004184, 0.40263037, 0.58500625, 0.47131007, 0.49528316, 0.65274651, 0.56246206, 0.59342141, 0.34467363, 0.5241102 , 0.53508769, 0.63755611, 0.54296967, 0.49532379, 0.55710514, 0.4727814 , 0.54287473, 0.54378878, 0.49598911, 0.44567613, 0.4226033 , 0.64046948, 0.424174  , 0.61146764, 0.53384126, 0.55580259, 0.53620641, 0.51944699, 0.44616688, 0.51671817, 0.47795645, 0.5214811 , 0.53330363, 0.43300768, 0.42334047, 0.35713199, 0.40525562, 0.56143502, 0.53251642, 0.57514168, 0.49649141, 0.66445602, 0.50344163, 0.58956719, 0.56268392, 0.52784727, 0.44697895, 0.54770617, 0.53614504, 0.48910506, 0.55732374, 0.6531372 , 0.45389849, 0.42714495, 0.50654265, 0.48575044, 0.54396153, 0.43507166, 0.49297671, 0.46060164, 0.33423935, 0.50641641, 0.51755618, 0.50710273, 0.58446034, 0.47326171, 0.56809047, 0.4116598 , 0.42240973, 0.51417129, 0.27841046, 0.45627347, 0.68837064, 0.46094653, 0.48798373, 0.6511203 , 0.55781644, 0.50132081, 0.68950118, 0.48040854, 0.4163646 , 0.44550726, 0.64579671, 0.55585583, 0.53346571, 0.54568866, 0.49305727, 0.5152283 , 0.61874024, 0.390471  , 0.57150632, 0.32169275, 0.56058687, 0.57962266, 0.42481201, 0.53675902, 0.58994957, 0.43245621, 0.52331344, 0.52329625, 0.43588796, 0.5770294 , 0.65409674, 0.52500875, 0.61645547, 0.46385359, 0.4056565 , 0.55277716, 0.51648013, 0.43605219, 0.45182095, 0.5321275 , 0.4447037 , 0.63397192, 0.62649506, 0.46668121, 0.4728771 , 0.52408038, 0.55460523, 0.49997893, 0.48603239, 0.47103121, 0.50781173, 0.6053177 , 0.49600114, 0.66377765, 0.41521762, 0.45647306, 0.60474739, 0.51181494, 0.48838955, 0.52510073, 0.61873979, 0.59586349, 0.57212219, 0.47198398, 0.5853836 , 0.42265219, 0.56646153, 0.4917164 , 0.57298278, 0.45851783, 0.39114968, 0.47425745, 0.50817601, 0.47754206, 0.67444964, 0.42658415, 0.42328879, 0.15379653, 0.51150188, 0.45829511, 0.42705282, 0.46390192, 0.52950714, 0.62715833, 0.58637768, 0.54803618, 0.49196057, 0.46279048, 0.50294143, 0.59226854, 0.60229713, 0.54225779, 0.5941689 , 0.47815737, 0.49883862, 0.48486635, 0.47261218, 0.43825047, 0.54336212, 0.51513842, 0.49567064, 0.61158689, 0.4129319 , 0.40661419, 0.51351605, 0.55734222, 0.51180806, 0.37477852, 0.49815911, 0.44054231, 0.53456185, 0.44864891, 0.58385036, 0.58104914, 0.59503922, 0.46185033, 0.44327255, 0.52114896, 0.61396954, 0.63000332, 0.66269052, 0.43450281, 0.39520424, 0.51656259, 0.45925242, 0.58154243, 0.6988516 , 0.4913753 , 0.4894337 , 0.56580656, 0.59821097, 0.5797311 , 0.41818885, 0.4976684 , 0.522177  , 0.68252638, 0.56276602, 0.37200166, 0.50755061, 0.3880415 , 0.49663169, 0.35715457, 0.54248505, 0.43303048, 0.64253914, 0.47356476, 0.51391693, 0.48855358, 0.37799246, 0.49398403, 0.56951529, 0.37688991, 0.49160708, 0.52690356, 0.52686674, 0.51215801, 0.57698389, 0.61241361, 0.52619258, 0.35055098, 0.48837412, 0.4407033 , 0.50077959, 0.45479108, 0.51414254, 0.63376111, 0.45150698, 0.46214344, 0.54575133, 0.60024207, 0.55953047, 0.51969718, 0.48523973, 0.47544462, 0.58340948, 0.38439877, 0.52759982, 0.41346489, 0.38991315, 0.68026434, 0.45986472, 0.43181916, 0.50648118, 0.3894343 , 0.62873892, 0.6592354 , 0.65293863, 0.46845371, 0.45012085, 0.57589279, 0.39502274, 0.52358881, 0.48640999, 0.54039256, 0.47576126, 0.62924319, 0.54273178, 0.50646707, 0.50287318, 0.59946252, 0.56940881, 0.59930375, 0.51774736, 0.48263752, 0.53941484, 0.49248521, 0.57155063, 0.4264775 , 0.51036859, 0.71054727, 0.53996883, 0.42901143, 0.41232303, 0.44549526, 0.38168336, 0.40328314, 0.53762153, 0.61349677, 0.60404878, 0.44697336, 0.54711955, 0.48216981, 0.55000164, 0.64022518, 0.52600931, 0.44678145, 0.43829697, 0.58384623, 0.43745937, 0.68318971, 0.45586054, 0.60599063, 0.4499793 , 0.42205188, 0.49386855, 0.5555232 , 0.5126009 , 0.3240312 , 0.55874947, 0.47750057, 0.58195061, 0.48240565, 0.48431244, 0.63907141, 0.42301053, 0.54758077, 0.64710017, 0.56611208, 0.3826722 , 0.51005073, 0.56860342, 0.46451647, 0.3152289 , 0.40651669, 0.54816232, 0.51171719, 0.61816035, 0.4923172 , 0.51395075, 0.4347128 , 0.6575245 , 0.43948447, 0.50860031, 0.52244094, 0.45780077, 0.5074908 , 0.5148614 , 0.63876247, 0.40235826, 0.45600197, 0.57247591, 0.58371552, 0.48872538, 0.56422632, 0.61667083, 0.47123427, 0.5550912 , 0.43315088, 0.50986655, 0.54065827, 0.44640175, 0.50459281, 0.54848554, 0.57166077, 0.47517982, 0.56219177, 0.50764385, 0.53088791, 0.43948164, 0.41238775, 0.70538166, 0.51623739, 0.56201525, 0.53010466, 0.50596836, 0.61114539, 0.42662833, 0.49634016, 0.46679557, 0.55409602, 0.38539755, 0.488031  , 0.57784212, 0.41440968, 0.69530549, 0.64241962, 0.43608626, 0.57265769, 0.41271096, 0.47754829, 0.57995155, 0.58981953, 0.42852998, 0.75993644, 0.36916639, 0.54238425, 0.55902348, 0.63663933, 0.56209095, 0.67578255, 0.47549209, 0.55657308, 0.60054174, 0.62168949, 0.45213496, 0.62205182, 0.50193585, 0.4553445 , 0.64101172, 0.65653997, 0.48887577, 0.50473288, 0.46740208, 0.60913356, 0.36652889, 0.53297661, 0.46235313, 0.50740879, 0.49457405, 0.59901308, 0.51292321, 0.42683503, 0.50625828, 0.4959117 , 0.50273795, 0.53775403, 0.56970437, 0.47405082, 0.54581842, 0.48832954, 0.49136057, 0.50149924, 0.54146932, 0.61667673, 0.55388527, 0.60724866, 0.55430024, 0.44857739, 0.55097295, 0.50442061, 0.50053846, 0.55487323, 0.50363833, 0.75506435, 0.35639168, 0.41038472, 0.55952598, 0.50743166, 0.49455597, 0.52371176, 0.5173707 , 0.4838424 , 0.65324038, 0.42345538, 0.63803063, 0.44717069, 0.45346519, 0.55144477, 0.32496316, 0.55187278, 0.38494981, 0.47160413, 0.54843355, 0.53055175, 0.50776486, 0.40104469, 0.58673325, 0.32903998, 0.67350565, 0.60284702, 0.53481366, 0.57032654, 0.59341211, 0.63042724, 0.46316767, 0.44645781, 0.50656859, 0.55613637, 0.51819588, 0.49919472, 0.69062766, 0.4478852 , 0.40277836, 0.44662984, 0.39562382, 0.46802174, 0.55769792, 0.49113326, 0.5389061 , 0.49699685, 0.4253304 , 0.49626194, 0.48616124, 0.46915155, 0.4410376 , 0.44304078, 0.57270865, 0.44673671, 0.6273081 , 0.369879  , 0.62204846, 0.44471064, 0.54359084, 0.52687983, 0.59604323, 0.61300378, 0.52645241, 0.53298558, 0.55336491, 0.43435126, 0.52870345, 0.46439409, 0.51879582, 0.3975725 , 0.54536605, 0.43020316, 0.68023616, 0.59403872, 0.6203501 , 0.62413604, 0.54910077, 0.46422285, 0.56620622, 0.54509271, 0.53949781, 0.53630026, 0.48322337, 0.55359606, 0.48841351, 0.44999576, 0.29540711, 0.36641105, 0.58706595, 0.58113201, 0.68880479, 0.55984352, 0.44740224, 0.45374018, 0.56251856, 0.49982744, 0.42061123, 0.540602  , 0.46889305, 0.44809521, 0.52279221, 0.60142552, 0.55545244, 0.51801667, 0.59008714, 0.50195991, 0.45056717, 0.39206303, 0.66207818, 0.54601705, 0.45171585, 0.63569838, 0.53855245, 0.6438954 , 0.440545  , 0.54951411, 0.59923302, 0.54129092, 0.56813283, 0.53986001, 0.41527197, 0.36456438, 0.61115221, 0.43983947, 0.73800103, 0.63074972, 0.49312482, 0.5173372 , 0.6329138 , 0.66093455, 0.50369289, 0.45906753, 0.50068213, 0.53136582, 0.5807486 , 0.43492245, 0.58062931, 0.48805741, 0.57597607, 0.51686055, 0.54750925, 0.48997315, 0.53736693, 0.62362939, 0.50190193, 0.5325812 , 0.46860223, 0.324969  , 0.68992969, 0.48199164, 0.53319143, 0.48993979, 0.45507132, 0.40559109, 0.5313514 , 0.53757575, 0.57961045, 0.64597933, 0.57550549, 0.45433565, 0.53549255, 0.49604919, 0.54922105, 0.44656997, 0.41898935, 0.46704811, 0.40511525, 0.57416591, 0.62893958, 0.53005504, 0.48683369, 0.49095719, 0.42303395, 0.5588482 , 0.44332021, 0.45659543, 0.488704  , 0.57998646, 0.61292463, 0.52526187, 0.43046489, 0.58988045, 0.43686196, 0.33904803, 0.37169298, 0.58140386, 0.59161022, 0.45171061, 0.57726815, 0.49122047, 0.55230574, 0.62887268, 0.6049496 , 0.45188227, 0.42203535, 0.46189681, 0.43825745, 0.44873449, 0.49085174, 0.50528202, 0.57305221, 0.42562782, 0.41511399, 0.43465258, 0.47183632, 0.448959  , 0.4758921 , 0.44645844, 0.56073963, 0.46712308, 0.38712153, 0.53403816, 0.50611533, 0.57749213, 0.4119324 , 0.38469673, 0.54295284, 0.42578656, 0.58108977, 0.63394487, 0.55756884, 0.44798806, 0.47436265, 0.45380543, 0.42918358, 0.42031482, 0.48113348, 0.57471113, 0.54689865, 0.55711705, 0.53265263, 0.36344944, 0.63173643, 0.51780289, 0.56665978, 0.5226248 , 0.46089314, 0.37125002, 0.31044164, 0.45929208, 0.60544915, 0.55353608, 0.36384265, 0.54001723, 0.45842086, 0.46196712, 0.63659072, 0.43846166, 0.55006328, 0.62314192, 0.65332482, 0.60704582, 0.50083931, 0.59155066, 0.6339093 , 0.56133137, 0.30652547, 0.38021809, 0.3768038 , 0.49119722, 0.48971265, 0.50806104, 0.47178806, 0.5725523 , 0.55252357, 0.44570776, 0.50177159, 0.48449303, 0.50532486, 0.46102534, 0.57731553, 0.50300858, 0.4994449 , 0.51758241, 0.49668101, 0.62251686, 0.44982344, 0.51133861, 0.43023747, 0.54125065, 0.53676388, 0.42292666, 0.43988441, 0.47708977, 0.50241534, 0.43362649, 0.57962462, 0.55244867, 0.52908234, 0.48711656, 0.55886961, 0.49564528, 0.40348658, 0.57285293, 0.39444   , 0.45864142, 0.57913263, 0.39257007, 0.49035964, 0.62788056, 0.47875583, 0.42241179, 0.60475894, 0.56127533, 0.46965246, 0.52615652, 0.41990849, 0.58733638, 0.48315409, 0.5611254 , 0.47869867, 0.50380561, 0.52108088, 0.57642078, 0.45127346, 0.54358445, 0.434648  , 0.62356916, 0.48690904, 0.59995824, 0.52712623, 0.45945608, 0.4492812 , 0.46278369, 0.3911091 , 0.62348611, 0.42424114, 0.5332421 , 0.48790962, 0.58100272, 0.61565383, 0.55114213, 0.480735  , 0.52528752, 0.49916826, 0.52733298, 0.47242066, 0.58371394, 0.51852892, 0.49704594, 0.3666496 , 0.52660887, 0.53144665, 0.5683481 , 0.36025109, 0.48266492, 0.43751574, 0.57474904, 0.64609875, 0.52933667, 0.35417038, 0.52756263, 0.46189389, 0.42695956, 0.46915893, 0.37413352, 0.46512703, 0.4969237 , 0.54969566, 0.63353521, 0.62107129, 0.50369703, 0.57078702, 0.38181758, 0.5265982 , 0.53785884, 0.48122966, 0.50826552, 0.66617047, 0.59056478, 0.70093275, 0.51854438, 0.51221269, 0.31020568, 0.50122871, 0.37391402, 0.56014951, 0.53847864, 0.53008847, 0.54913291, 0.51045918, 0.49561606, 0.64418362, 0.53500265, 0.49586041, 0.52483259, 0.38958657, 0.62136111, 0.42181536, 0.69896511, 0.5410416 , 0.55431738, 0.52601822, 0.62970176, 0.63491201, 0.54844052, 0.43744364, 0.41544044, 0.54385721, 0.5441108 , 0.50304093, 0.55859643, 0.45248808, 0.53529228, 0.51971409, 0.53501217, 0.58076224, 0.57613014, 0.48832413, 0.66410428, 0.63485361, 0.43496978, 0.50512161, 0.50838357, 0.54934211, 0.48391504, 0.48480158, 0.43940921, 0.39948247, 0.63931175, 0.57068969, 0.56312944, 0.38171528, 0.53457397, 0.60110379, 0.48242732, 0.49683279, 0.57717104, 0.54168638, 0.53981359, 0.53076083, 0.40198876, 0.43509165, 0.48128789, 0.50833319, 0.5022883 , 0.49263786, 0.44034847, 0.38583003, 0.5257617 , 0.33333216, 0.57228046, 0.44311679, 0.50628454, 0.54734323, 0.61508466, 0.4089733 , 0.29435893, 0.48159636, 0.44519163, 0.58689508, 0.43574508, 0.576658  , 0.45179824, 0.5514395 , 0.55504375, 0.4668765 , 0.4518658 , 0.4687282 , 0.34070807, 0.55836204, 0.56475398, 0.51378136, 0.55962147, 0.45479384, 0.43671538, 0.60970039, 0.40858464, 0.43676123, 0.70214932, 0.44041219, 0.47710043, 0.47999604, 0.54908587, 0.74167159, 0.53091758, 0.44484116, 0.54782792, 0.58267909])

# to run the full simulation uncomment the following line to fit the model for every dataset and not just for the first dataset
#for i_rep in range(n_rep):
In [42]: for i_rep in range(1):
....:     (x, y, d) = data[i_rep]
....:     obj_dml_data = DoubleMLData.from_arrays(x, y, d)
....:     obj_dml_plr = DoubleMLPLR(obj_dml_data,
....:                               ml_g, ml_m,
....:                               n_folds=2,
....:                               score='IV-type')
....:     obj_dml_plr.fit()
....:     this_theta = obj_dml_plr.coef[0]
....:     print(np.abs(theta_dml[i_rep] - this_theta))
....:     theta_dml[i_rep] = this_theta
....:
3.678763647452232e-05

In [43]: ax = sns.kdeplot(theta_dml, shade=True, color=colors[3])

In [44]: ax.axvline(0.5, color='k', label='True theta');

set.seed(3333)

# to speed up the illustration we hard-code the simulation results
theta_dml = c(0.42913806, 0.43316824, 0.61369310, 0.47420288, 0.56612731, 0.48624066, 0.48266824, 0.42687167, 0.57932496, 0.35014429, 0.45364242, 0.50362907, 0.35426940, 0.50631755, 0.69905326, 0.44184129, 0.68664042, 0.65077473, 0.54085236, 0.53393103, 0.41080641, 0.55493858, 0.38648321, 0.50326788, 0.47792825, 0.52250731, 0.45095439, 0.51617920, 0.72629732, 0.44585236, 0.57614603, 0.40003120, 0.56031886, 0.51314163, 0.51161077, 0.42458006, 0.45470411, 0.62995282, 0.52821749, 0.52106290, 0.41796137, 0.49244195, 0.44735716, 0.47464688, 0.43278999, 0.57738025, 0.46497964, 0.51932182, 0.41720577, 0.43764400, 0.67841266, 0.55408659, 0.51307914, 0.68270399, 0.59202605, 0.56587065, 0.54677982, 0.53995167, 0.40541990, 0.43919669, 0.41382584, 0.60354425, 0.47532433, 0.57520404, 0.34593387, 0.44649214, 0.47314742, 0.65742616, 0.57004944, 0.58934731, 0.57040034, 0.64863370, 0.55508407, 0.56819856, 0.52013619, 0.46425486, 0.62237868, 0.51925107, 0.52501949, 0.62183034, 0.51362139, 0.50258909, 0.53324487, 0.37252394, 0.50511082, 0.56329268, 0.46591064, 0.46891112, 0.36117298, 0.40960975, 0.56476302, 0.60346171, 0.55011963, 0.43666665, 0.43957066, 0.59957770, 0.49008623, 0.67188050, 0.57097026, 0.60687472, 0.60862364, 0.50827286, 0.50359364, 0.30725406, 0.40085167, 0.50973296, 0.55199812, 0.49537286, 0.45188790, 0.56682157, 0.60075966, 0.55716112, 0.36188043, 0.42720108, 0.48913582, 0.49905466, 0.30636339, 0.37399033, 0.47361759, 0.55287262, 0.67421189, 0.65452966, 0.70761480, 0.41981470, 0.48624085, 0.46167888, 0.45667825, 0.47723405, 0.45471222, 0.47466022, 0.52806844, 0.55456203, 0.35834979, 0.48054982, 0.40626663, 0.53222057, 0.50406679, 0.61914153, 0.45274682, 0.38146984, 0.55040314, 0.51683918, 0.62819509, 0.56447129, 0.40844310, 0.56361326, 0.59545085, 0.54341239, 0.38600811, 0.48514795, 0.61624648, 0.52223771, 0.50694287, 0.40784649, 0.43925088, 0.56120925, 0.53709777, 0.53361594, 0.60162763, 0.55301727, 0.57168732, 0.40383638, 0.61029386, 0.32967098, 0.44418344, 0.59388784, 0.48403921, 0.43563524, 0.57586206, 0.54438640, 0.39391988, 0.54026644, 0.50304476, 0.45330140, 0.51772101, 0.57072679, 0.47129530, 0.52952563, 0.51118177, 0.43792613, 0.50708501, 0.55402204, 0.56300018, 0.64710173, 0.61068413, 0.50885259, 0.46867644, 0.51529826, 0.54432131, 0.44366517, 0.47907430, 0.48360119, 0.46415220, 0.62803625, 0.51681093, 0.44861562, 0.46878897, 0.57754910, 0.53934414, 0.55514825, 0.64211494, 0.58543153, 0.56421745, 0.41812980, 0.61842351, 0.39951353, 0.65326039, 0.59645961, 0.49958845, 0.39999194, 0.48458269, 0.57670828, 0.48991088, 0.46882587, 0.56076353, 0.45412254, 0.57411999, 0.44973420, 0.38673987, 0.49782242, 0.60652263, 0.32827752, 0.40505171, 0.52913411, 0.58860288, 0.54656647, 0.59747115, 0.50592872, 0.60282387, 0.34342820, 0.38265266, 0.37658983, 0.46327940, 0.45987841, 0.47207120, 0.62249832, 0.49486043, 0.48560421, 0.42329878, 0.43797319, 0.53221423, 0.49708016, 0.39947860, 0.37220067, 0.62247380, 0.51071511, 0.73578123, 0.49977786, 0.47002894, 0.55606379,
0.60036521, 0.37144821, 0.56061871, 0.58470561, 0.52377130, 0.47195886, 0.46082198, 0.54618085, 0.51661456, 0.40254820, 0.79526518, 0.47619727, 0.64471049, 0.48303118, 0.50388135, 0.34803997, 0.35489409, 0.48573732, 0.56607696, 0.55677069, 0.66452679, 0.47796660, 0.36815952, 0.49030185, 0.46802058, 0.49364717, 0.53806177, 0.63226813, 0.41949031, 0.61031602, 0.52203176, 0.45365217, 0.52340408, 0.38941021, 0.41657998, 0.46790560, 0.51640168, 0.52023209, 0.41658159, 0.62603226, 0.64976888, 0.55140079, 0.40168700, 0.48450186, 0.37489314, 0.57126164, 0.42998898, 0.46747086, 0.46207941, 0.42358173, 0.40508553, 0.60585155, 0.55732323, 0.49555698, 0.52873767, 0.66544607, 0.42496742, 0.47372355, 0.45987312, 0.56542595, 0.55594871, 0.39854807, 0.47535006, 0.47627848, 0.50834030, 0.53412860, 0.35496264, 0.38224096, 0.53455446, 0.53345455, 0.51537376, 0.56669332, 0.61729402, 0.56773772, 0.51540255, 0.46143092, 0.46263424, 0.47306752, 0.71764595, 0.44831611, 0.44472624, 0.54153129, 0.45992811, 0.41468665, 0.50693472, 0.53712667, 0.50115649, 0.41561605, 0.39845802, 0.57098662, 0.47668312, 0.53814574, 0.54558775, 0.36989976, 0.57058142, 0.59245627, 0.39453265, 0.55638448, 0.41792649, 0.50174815, 0.42716142, 0.57053877, 0.50450045, 0.50719061, 0.40791363, 0.60117304, 0.50726098, 0.41313999, 0.62713006, 0.52927468, 0.43919127, 0.51939909, 0.53216099, 0.61426638, 0.55517891, 0.52550725, 0.48726381, 0.52282488, 0.53412265, 0.57112127, 0.44915620, 0.48624981, 0.45434757, 0.54594964, 0.45144261, 0.58943593, 0.44173928, 0.39137041, 0.56532290, 0.52862947, 0.59335260, 0.47949536, 0.59270141, 0.60803178, 0.44201101, 0.50200628, 0.57736589, 0.67365436, 0.47965685, 0.53265149, 0.29539617, 0.63942053, 0.36842280, 0.60947176, 0.52338164, 0.59342733, 0.49698428, 0.69892615, 0.63637029, 0.47600661, 0.50267946, 0.52290767, 0.51826972, 0.53976759, 0.49725101, 0.51986060, 0.55680582, 0.53644020, 0.56394591, 0.41015781, 0.56109820, 0.54985548, 0.48015589, 0.70643971, 0.50029105, 0.57595621, 0.51223989, 0.62096412, 0.44081326, 0.39901139, 0.55241395, 0.51862525, 0.57115813, 0.44345708, 0.54100182, 0.51750750, 0.51221937, 0.44313306, 0.46866454, 0.54853286, 0.45596388, 0.60353598, 0.59485756, 0.51394737, 0.52571069, 0.51710424, 0.45414012, 0.37452561, 0.48179420, 0.48400926, 0.54266013, 0.62673309, 0.51523876, 0.70196700, 0.70896770, 0.46851214, 0.59934315, 0.50147654, 0.49531880, 0.50256030, 0.63738873, 0.44100607, 0.50199621, 0.50267163, 0.57522836, 0.54297195, 0.59369577, 0.47387198, 0.46236344, 0.55472654, 0.64741476, 0.56064831, 0.48637194, 0.48030544, 0.55182318, 0.59873158, 0.38704244, 0.49207536, 0.61636373, 0.49309616, 0.43057533, 0.42656903, 0.57737717, 0.47652448, 0.44508640, 0.55721658, 0.42001502, 0.53203807, 0.67786061, 0.56307312, 0.59090498, 0.46763688, 0.60226755, 0.51045486, 0.58261018, 0.51335751, 0.51489954, 0.49543312, 0.54261570, 0.37144113, 0.56424704, 0.52789051, 0.62629321, 0.56680133, 0.58094908, 0.56792576, 0.38076462, 0.51025326, 0.41766390, 0.67992174,
0.57048006, 0.55394699, 0.57142484, 0.50051587, 0.53052551, 0.59685925, 0.57857539, 0.66474049, 0.40036740, 0.45197943, 0.48545635, 0.58707913, 0.52364072, 0.53642783, 0.50616922, 0.38669512, 0.53221757, 0.44462522, 0.56490506, 0.52281451, 0.61329588, 0.53618070, 0.54026911, 0.57958907, 0.44172998, 0.54616971, 0.49559402, 0.46337158, 0.40539251, 0.48660211, 0.52950782, 0.67319498, 0.50960922, 0.55025810, 0.48466318, 0.48733741, 0.57079932, 0.52210722, 0.59483531, 0.57823919, 0.56681829, 0.67063110, 0.55044265, 0.49411047, 0.49592631, 0.53839330, 0.50303196, 0.57088299, 0.53512485, 0.45772366, 0.52151506, 0.58684530, 0.41629452, 0.48724706, 0.49347072, 0.44468953, 0.62079071, 0.45174209, 0.50964193, 0.57956285, 0.46846437, 0.45610451, 0.46687298, 0.56776141, 0.50069025, 0.47016823, 0.52769089, 0.48033467, 0.47567157, 0.51580042, 0.54601488, 0.44649238, 0.52159010, 0.58909856, 0.44350571, 0.47345207, 0.47027266, 0.52040233, 0.56983877, 0.47086448, 0.53872263, 0.44996677, 0.47074500, 0.42960621, 0.52333355, 0.58375512, 0.46329712, 0.49498312, 0.59458654, 0.50953637, 0.63728315, 0.50204231, 0.59437553, 0.52019228, 0.47574961, 0.43201755, 0.44222095, 0.53067411, 0.55695602, 0.40546530, 0.51287913, 0.52125397, 0.42011085, 0.44796920, 0.66000531, 0.55670752, 0.47266941, 0.49879186, 0.57275709, 0.46730920, 0.61246393, 0.56723568, 0.61656772, 0.53631019, 0.57331834, 0.57269018, 0.46993462, 0.49854259, 0.39883510, 0.57493417, 0.51821134, 0.43730615, 0.54825472, 0.64819963, 0.48815339, 0.47117569, 0.58801247, 0.38179558, 0.49153756, 0.53372616, 0.39952425, 0.54495609, 0.56425681, 0.60914698, 0.67661682, 0.43440964, 0.47919319, 0.50718258, 0.51064880, 0.73991197, 0.58496491, 0.59005490, 0.46279519, 0.58139833, 0.48246194, 0.47157441, 0.51921259, 0.59101700, 0.40841472, 0.57875099, 0.40205734, 0.52224831, 0.51818062, 0.53313627, 0.61172011, 0.49113850, 0.43693161, 0.46919091, 0.63009573, 0.55165641, 0.64606963, 0.55488746, 0.54408366, 0.38395325, 0.55924915, 0.51209453, 0.39222947, 0.44400452, 0.45858189, 0.50094515, 0.54827255, 0.61142796, 0.50540355, 0.50289919, 0.52269712, 0.44278828, 0.54354751, 0.40389553, 0.48896288, 0.51261510, 0.59240684, 0.57118052, 0.51133825, 0.70668058, 0.55364559, 0.56129073, 0.42279244, 0.51104222, 0.50439600, 0.40569372, 0.47728061, 0.43199642, 0.70508883, 0.41362643, 0.54311643, 0.44714492, 0.42709630, 0.37506699, 0.47275717, 0.65199031, 0.55824739, 0.47710929, 0.52287265, 0.46701169, 0.34957259, 0.57366553, 0.61544005, 0.57923248, 0.48491284, 0.67073890, 0.59296263, 0.52960127, 0.48830529, 0.42371237, 0.53493779, 0.46803035, 0.35813584, 0.63742714, 0.53100854, 0.48004508, 0.38527609, 0.45555077, 0.61923367, 0.45278751, 0.55398663, 0.54611470, 0.47799756, 0.62941151, 0.53417734, 0.48169190, 0.57393957, 0.56965073, 0.48438376, 0.59291956, 0.58419322, 0.40210205, 0.40705906, 0.60354322, 0.66776203, 0.47749758, 0.51099112, 0.41384870, 0.50893890, 0.44254692, 0.52634243, 0.48099904, 0.58424289, 0.59455376, 0.69857315, 0.52939985,
0.51870330, 0.52861707, 0.56935708, 0.65242725, 0.33287675, 0.36760792, 0.44377420, 0.56422963, 0.60120575, 0.41819252, 0.56997243, 0.60729652, 0.53456727, 0.49915985, 0.49840749, 0.54267287, 0.54345681, 0.51760500, 0.36975586, 0.53025333, 0.66200912, 0.50074647, 0.40370975, 0.50420793, 0.44299858, 0.58169330, 0.37495283, 0.35753936, 0.61401972, 0.38445478, 0.66316660, 0.39712808, 0.60481713, 0.43205297, 0.60673487, 0.36795722, 0.58504703, 0.59312592, 0.50726682, 0.43250603, 0.59993261, 0.50850652, 0.55197866, 0.42100227, 0.41197278, 0.48603494, 0.48233252, 0.52166479, 0.48123320, 0.57925253, 0.45858549, 0.44852927, 0.63552822, 0.67505307, 0.64072903, 0.58062243, 0.53493671, 0.57334320, 0.41484093, 0.50335731, 0.47572475, 0.62807268, 0.51276249, 0.48341935, 0.57002321, 0.42448958, 0.46463977, 0.50158529, 0.50336207, 0.38493723, 0.49877225, 0.70920972, 0.52804881, 0.55241536, 0.51123088, 0.56935502, 0.46794885, 0.62160760, 0.47682064, 0.55061002, 0.45601717, 0.32226202, 0.50512015, 0.39238552, 0.51890464, 0.56729508, 0.48812630, 0.52579096, 0.49552910, 0.45541965, 0.57388623, 0.55532925, 0.42905776, 0.42528833, 0.50058720, 0.57588967, 0.58886503, 0.45381248, 0.50284578, 0.44971131, 0.55089234, 0.50441897, 0.46913765, 0.49684019, 0.48161077, 0.55911804, 0.50199758, 0.47945146, 0.66388549, 0.39635503, 0.53354409, 0.64561425, 0.66166898, 0.47160421, 0.50143847, 0.64010436, 0.46946532, 0.52131207, 0.53527948, 0.47010874, 0.51834605, 0.54124221, 0.40402007, 0.50168993, 0.50234726, 0.40501895, 0.47435409, 0.39967228, 0.58627553, 0.63872443, 0.51361588, 0.47086834, 0.44200927, 0.54694174, 0.53323022, 0.34121376, 0.46739575, 0.42969545, 0.70104546, 0.46003349, 0.51258474, 0.72598733, 0.50390373, 0.50638444, 0.50122981, 0.51856086, 0.58824995, 0.56395614, 0.58458643, 0.48689805, 0.54648130, 0.46113281, 0.42569036, 0.51534597, 0.42647065, 0.36155963, 0.39515078, 0.59939998, 0.49517795, 0.43462746, 0.47664938, 0.56940909, 0.57171369, 0.60772723, 0.48845162, 0.64587916, 0.51106531, 0.56538809, 0.61896522, 0.45928391, 0.55184185, 0.59693194, 0.59348120, 0.43120609, 0.54165183, 0.31393692, 0.74020994, 0.58668672, 0.44618824, 0.58408901, 0.50309948, 0.50841443, 0.51478630, 0.48087840, 0.51480573, 0.50754975, 0.46715053, 0.58687497, 0.57562903, 0.56412152, 0.49320614, 0.48217559, 0.45551353, 0.50503672, 0.44996034, 0.63371039, 0.51740116, 0.41814389, 0.47413265, 0.55776780, 0.45872666, 0.59620842, 0.65898842, 0.65894305, 0.56710870, 0.60647653, 0.54772268, 0.42489299, 0.58600208, 0.33854023, 0.35980618, 0.43478661, 0.56196962, 0.60073501, 0.43790578, 0.39961190, 0.50526949, 0.49429539, 0.42253674, 0.42425078, 0.62641974, 0.43744788, 0.28745392, 0.49301899, 0.54976290, 0.51968761, 0.51677942, 0.52912944, 0.47229849, 0.47875525, 0.62221667, 0.58038527, 0.45193207, 0.47508698, 0.43425117, 0.45337453, 0.50681186, 0.45026174, 0.55684403, 0.39043084, 0.45808880, 0.39807946, 0.67317732, 0.48568494, 0.50180223, 0.61079194, 0.58155068, 0.55683503, 0.53079425, 0.44263915)

# to run the full simulation uncomment the following line to fit the model for every dataset and not just for the first dataset
#for (i_rep in seq_len(n_rep)) {
for (i_rep in seq_len(1)) {
df = data[[i_rep]]
obj_dml_data = double_ml_data_from_data_frame(df, y_col = "y", d_cols = "d")
obj_dml_plr = DoubleMLPLR$new(obj_dml_data, ml_g, ml_m, n_folds=2, score='IV-type') obj_dml_plr$fit()
this_theta = obj_dml_plr$coef print(abs(theta_dml[i_rep] - this_theta)) theta_dml[i_rep] = this_theta } g_dml = ggplot(data.frame(theta_dml), aes(x = theta_dml)) + geom_density(fill = "dark red", alpha = 0.3, color = "dark red") + geom_vline(aes(xintercept = alpha), col = "black") + xlim(c(0.08, 0.75)) + xlab("") + ylab("") + theme_minimal() g_dml   d 3.479779e-09  ## 1.6. Double/debiased machine learning¶ To illustrate the benefits of the auxiliary prediction step in the DML framework we write the error as $\sqrt{n}(\check{\theta} - \theta) = a^* + b^* + c^*$ Chernozhukov et al. (2018) argues that: The first term $a^* := (EV^2)^{-1} \frac{1}{\sqrt{n}} \sum_{i\in I} V_i U_i$ will be asymptotically normally distributed. The second term $b^* := (EV^2)^{-1} \frac{1}{\sqrt{n}} \sum_{i\in I} (\hat{m}(X_i) - m(X_i)) (\hat{g}_0(X_i) - g_0(X_i))$ vanishes asymptotically for many data generating processes. The third term $$c^*$$ vanishes in probability if sample splitting is applied. In [45]: ax = sns.kdeplot(theta_ols, shade=True) In [46]: sns.kdeplot(theta_nonorth, shade=True, ax=ax); In [47]: sns.kdeplot(theta_orth_nosplit, shade=True); In [48]: sns.kdeplot(theta_dml, shade=True); In [49]: labels = ['True$\\theta\$', 'OLS', 'Non-orthogonal ML', 'Double ML (no sample splitting)', 'Double ML with cross-fitting']

In [50]: ax.axvline(0.5, color='k', label='True theta');

In [51]: ax.legend(labels);

g_all = ggplot(data.frame(theta_ols, theta_nonorth, theta_orth_nosplit, theta_dml)) +
geom_density(aes(x = theta_ols), fill = "dark blue", alpha = 0.3, color = "dark blue") +
geom_density(aes(x = theta_nonorth), fill = "dark orange", alpha = 0.3, color = "dark orange") +
geom_density(aes(x = theta_orth_nosplit), fill = "dark green", alpha = 0.3, color = "dark green") +
geom_density(aes(x = theta_dml), fill = "dark red", alpha = 0.3, color = "dark red") +
geom_vline(aes(xintercept = alpha), col = "black") +
xlim(c(0.08, 0.75)) + xlab("") + ylab("") + theme_minimal()
g_all


## 1.7. References¶

Chernozhukov, V., Chetverikov, D., Demirer, M., Duflo, E., Hansen, C., Newey, W. and Robins, J. (2018), Double/debiased machine learning for treatment and structural parameters. The Econometrics Journal, 21: C1-C68. doi:10.1111/ectj.12097.