effect_modifiers

[wxl112]

Describe the bug I'm trying to estimate the causal effect by calling the "econml" package and specified an "effect_modifiers" variables that is continuous value ,but also I want set the parameters "num_quantiles_to_discretize_cont_cols = 6"

Steps to reproduce the behavior

dml_estimate = model.estimate_effect(identified_estimand, method_name="backdoor.econml.dml.DML",
                                     control_value = 0,
                                     treatment_value = 1000,
                                     effect_modifiers=['solar_radiation'],
                                 target_units = 'ate',#lambda df: df["diffusion_conditions"]==1,  # condition used for CATE
                                 confidence_intervals=False,
                                     num_quantiles_to_discretize_cont_cols = 6,
                                method_params={"init_params":{
                                                              'model_y':GradientBoostingRegressor(),
                                                              'model_t': GradientBoostingRegressor(),
                                                              "model_final": LassoCV(fit_intercept=False),
                                                              'featurizer':PolynomialFeatures(degree=1, include_bias=True)
                                                              },
                                                "fit_params":{
                                                    'inference':BootstrapInference(n_bootstrap_samples=100, n_jobs=-1),}
                                               })

TypeError Traceback (most recent call last) ~\AppData\Local\Temp\ipykernel_3380\140799330.py in 25 }, 26 "fit_params":{ ---> 27 'inference':BootstrapInference(n_bootstrap_samples=100, n_jobs=-1),} 28 }) 29 print(dml_estimate)

TypeError: estimate_effect() got an unexpected keyword argument 'num_quantiles_to_discretize_cont_cols'

Expected behavior I really expected i would have report behavior like below:

Version information:

• DoWhy version [e.g. 0.7]

Additional context Add any other context about the problem here.

[amit-sharma]

Thanks for raising this @wxl112 . We do not currently support this parameter for econml estimators. Will plan to add it for the next release. For now, you may discretize the continuous effect modifiers before passing the data to DoWhy.

[wxl112]

thanks for taking your time to answer the question！ all the best！regard！

[Padarn]

It looks to me that this would work as long as the parameter is provided as extra_args? The EconML estimator: https://github.com/py-why/dowhy/blob/97e6bdc3db137280fdb8812dfba34de14a248c72/dowhy/causal_model.py#L307-L313 does have this as an argument: https://github.com/py-why/dowhy/blob/97e6bdc3db137280fdb8812dfba34de14a248c72/dowhy/causal_estimators/linear_regression_estimator.py#L32