Caml Synthetic Data API Usage

from caml.extensions.synthetic_data import CamlSyntheticDataGenerator

Generate Data

data =  CamlSyntheticDataGenerator(n_obs=10_000,
                                  n_cont_outcomes=1,
                                  n_binary_outcomes=1,
                                  n_cont_treatments=1,
                                  n_binary_treatments=1,
                                  n_discrete_treatments=1,
                                  n_cont_confounders=1,
                                  n_binary_confounders=1,
                                  n_discrete_confounders=1,
                                  n_cont_modifiers=1,
                                  n_binary_modifiers=1,
                                  n_discrete_modifiers=1,
                                  n_confounding_modifiers=1,
                                  stddev_outcome_noise=3,
                                  stddev_treatment_noise=3,
                                  causal_model_functional_form="linear",
                                  n_nonlinear_transformations=10,
                                  n_nonlinear_interactions=5,
                                  seed=15)

Simulated Dataframe

data.df

	W1_continuous	W2_binary	W3_discrete	X1_continuous	X2_binary	X3_discrete	T1_continuous	T2_binary	T3_discrete	Y1_continuous	Y2_binary
0	0.706567	1	0	-1.652559	1	1	2.494731	1	1	-25.198250	1
1	3.315282	1	2	-2.775285	1	0	7.464238	0	0	-70.792577	1
2	1.908430	1	3	-3.451824	0	1	3.470709	0	1	-43.344150	1
3	3.743169	1	0	-4.319313	1	0	8.065827	0	4	-131.063425	1
4	2.781994	1	3	-2.901003	1	0	8.161921	0	4	-107.833216	1
...	...	...	...	...	...	...	...	...	...	...	...
9995	1.853102	0	2	-2.369628	0	1	4.856855	0	0	-30.105565	1
9996	1.710607	1	2	-2.642743	1	1	4.286633	0	0	-45.993240	1
9997	4.942256	0	0	-2.908543	0	0	12.208850	0	4	-103.391312	1
9998	3.125018	1	1	-3.039957	0	3	15.905126	0	4	-165.097366	0
9999	2.500979	0	0	-3.320707	0	3	10.460442	0	4	-127.790919	1

10000 rows × 11 columns

DGP

for k,v in data.dgp.items():
    print(k)
    print(v)

T1_continuous
      covariates    params global_transformation
0  W1_continuous  2.601096                  None
1      W2_binary -1.214932                  None
2    W3_discrete -0.128753                  None
3    X3_discrete  1.605966                  None
T2_binary
      covariates    params global_transformation
0  W1_continuous -1.633325               Sigmoid
1      W2_binary  0.331532               Sigmoid
2    W3_discrete -2.018670               Sigmoid
3    X3_discrete  0.281986               Sigmoid
T3_discrete
      covariates  cat_1_params  cat_2_params global_transformation
0  W1_continuous      1.251009     -2.752482               Softmax
1      W2_binary     -0.364808      2.056047               Softmax
2    W3_discrete     -0.068131      1.673660               Softmax
3    X3_discrete      1.310578      2.357792               Softmax
Y1_continuous
                         covariates    params global_transformation
0                     W1_continuous -1.553833                  None
1                         W2_binary -2.998091                  None
2                       W3_discrete -2.915301                  None
3                     X1_continuous -1.813285                  None
4                         X2_binary  0.739353                  None
5                       X3_discrete  0.082275                  None
6                     T1_continuous -1.643250                  None
7                         T2_binary -0.409790                  None
8                       T3_discrete -2.571727                  None
9   int_T1_continuous_X1_continuous  1.374265                  None
10      int_T1_continuous_X2_binary -2.330711                  None
11    int_T1_continuous_X3_discrete -0.588478                  None
12      int_T2_binary_X1_continuous -1.927494                  None
13          int_T2_binary_X2_binary -0.416015                  None
14        int_T2_binary_X3_discrete -1.346297                  None
15    int_T3_discrete_X1_continuous  1.974255                  None
16        int_T3_discrete_X2_binary -0.304700                  None
17      int_T3_discrete_X3_discrete -0.754942                  None
Y2_binary
                         covariates    params global_transformation
0                     W1_continuous -0.906427               Sigmoid
1                         W2_binary -2.316021               Sigmoid
2                       W3_discrete  0.545442               Sigmoid
3                     X1_continuous -0.204994               Sigmoid
4                         X2_binary  0.417902               Sigmoid
5                       X3_discrete  2.195866               Sigmoid
6                     T1_continuous  2.169852               Sigmoid
7                         T2_binary -2.210767               Sigmoid
8                       T3_discrete -2.387556               Sigmoid
9   int_T1_continuous_X1_continuous -0.145369               Sigmoid
10      int_T1_continuous_X2_binary -1.010015               Sigmoid
11    int_T1_continuous_X3_discrete -1.460272               Sigmoid
12      int_T2_binary_X1_continuous  1.614527               Sigmoid
13          int_T2_binary_X2_binary  2.939611               Sigmoid
14        int_T2_binary_X3_discrete -0.418421               Sigmoid
15    int_T3_discrete_X1_continuous -2.972694               Sigmoid
16        int_T3_discrete_X2_binary  2.586574               Sigmoid
17      int_T3_discrete_X3_discrete -1.248249               Sigmoid

True Conditional Average Treatment Effects (CATEs)

data.cates

	CATE_of_T1_continuous_on_Y1_continuous	CATE_of_T2_binary_on_Y1_continuous	CATE_of_T3_discrete_on_Y1_continuous_level_1_v_0	CATE_of_T3_discrete_on_Y1_continuous_level_4_v_0	CATE_of_T3_discrete_on_Y1_continuous_level_3_v_0	CATE_of_T3_discrete_on_Y1_continuous_level_2_v_0	CATE_of_T1_continuous_on_Y2_binary	CATE_of_T2_binary_on_Y2_binary	CATE_of_T3_discrete_on_Y2_binary_level_1_v_0	CATE_of_T3_discrete_on_Y2_binary_level_4_v_0	CATE_of_T3_discrete_on_Y2_binary_level_3_v_0	CATE_of_T3_discrete_on_Y2_binary_level_2_v_0
0	-6.833492	1.013194	-6.893941	-27.575765	-20.681824	-13.787882	0.000000	0.000000	0.166471	0.166471	0.166471	0.166471
1	-7.787938	4.523539	-8.355547	-33.422188	-25.066641	-16.711094	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
2	-6.975448	4.897281	-10.141449	-40.565795	-30.424346	-20.282897	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
3	-9.909840	7.499642	-11.403851	-45.615403	-34.211552	-22.807702	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
4	-7.960707	4.765859	-8.603745	-34.414982	-25.811236	-17.207491	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
...	...	...	...	...	...	...	...	...	...	...	...	...
9995	-5.488225	2.811356	-8.004919	-32.019678	-24.014758	-16.009839	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
9996	-8.194268	2.921768	-8.848818	-35.395272	-26.546454	-17.697636	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
9997	-5.640359	5.196408	-8.313932	-33.255728	-24.941796	-16.627864	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
9998	-7.586390	1.410815	-10.838203	-43.352811	-32.514608	-21.676405	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
9999	-7.972215	1.951960	-11.392476	-45.569903	-34.177427	-22.784951	-0.347813	-0.800617	0.000000	0.800617	0.069235	0.000000

10000 rows × 12 columns

True Average Treatment Effects (ATEs)

data.ates

	Treatment	ATE
0	T1_continuous_on_Y1_continuous	-7.420280
1	T2_binary_on_Y1_continuous	3.120486
2	T3_discrete_on_Y1_continuous_level_1_v_0	-9.381688
3	T3_discrete_on_Y1_continuous_level_4_v_0	-37.526754
4	T3_discrete_on_Y1_continuous_level_3_v_0	-28.145065
5	T3_discrete_on_Y1_continuous_level_2_v_0	-18.763377
6	T1_continuous_on_Y2_binary	-0.000474
7	T2_binary_on_Y2_binary	-0.109855
8	T3_discrete_on_Y2_binary_level_1_v_0	0.139524
9	T3_discrete_on_Y2_binary_level_4_v_0	0.247014
10	T3_discrete_on_Y2_binary_level_3_v_0	0.219988
11	T3_discrete_on_Y2_binary_level_2_v_0	0.186775