dame_flame.utils.data

Functions that generate data according to a distribution, returning the data and the true treatment effects of the values.

dame_flame.utils.data.generate_uniform_given_importance

Creates a dataset with covariates in a uniform distribution where the covariates’ importance can be pre-specified by the user. The treatment effect will be normally distributed.

generate_uniform_given_importance(num_control=1000, num_treated=1000,
                                  num_cov=4, min_val=0,
                                  max_val=3, covar_importance=[4, 3, 2, 1],
                                  bi_mean=2, bi_stdev=1)

Source Code

Parameter Name	Type	Description
num_control	integer	The number of units in the control group
num_treated	integer	The number of units in the treated group
num_cov	integer	The number of covariates
min_val	integer	The minimum value each covariate can take
max_val	integer	The maximum value each covariate can take
covar_importance	array	The importance each covariate has in determining the outcome. Must be of length num_cov
bi_mean	numerical	The treatment effect is normally distributed with this mean
bi_stdev	numerical	The treatment effect is normally distributed with this standard deviation

Return Name	Type	Description
data_frame	Pandas Dataframe	A dataframe that is according to the specifications. Units 0 through num_treated are treatment group, and num_treated through num_treated+num_control are control group units.
true_catt	array	The CATE of each treated unit, where the ith has the treatment effect of the ith unit in the dataframe.

dame_flame.utils.data.generate_binomial_given_importance

Creates a dataset with covariates in a binomial distribution where the covariates’ importance on outcome can be pre-specified by the user. The treatment effect will be normally distributed.

generate_binomial_given_importance(num_control=1000, num_treated=1000,
                                  num_cov=5, bernoulli_param=0.5,
                                  bi_mean=2, bi_stdev=1,
                                  covar_importance=[4, 3, 2, 1, 0.01])

Source Code

Parameter Name	Type	Description
num_control	integer	The number of units in the control group
num_treated	integer	The number of units in the treated group
num_cov	integer	The number of covariates
bernoulli_param	numerical	The bernoulli parameter determining the distribution of the covariates
covar_importance	array	The importance each covariate has in determining the outcome. Must be of length num_cov
bi_mean	numerical	The treatment effect is normally distributed with this mean
bi_stdev	numerical	The treatment effect is normally distributed with this standard deviation

Return Name	Type	Description
data_frame	Pandas Dataframe	A dataframe that is according to the specifications. Units 0 through num_treated are treatment group, and num_treated through num_treated+num_control are control group units.
true_catt	array	The CATE of each treated unit, where the ith has the treatment effect of the ith unit in the dataframe.

dame_flame.utils.data.generate_binomial_decay_importance

Creates a dataset with covariates in a binomial distribution where the covariates’ importance on outcome exponentially decays rather than being determined by the user. The treatment effect will be normally distributed.

generate_binomial_decay_importance(num_control=1000, num_treated=1000,
                                  num_cov=5, bernoulli_param=0.5,
                                  bi_mean=2, bi_stdev=1)

Source Code

Parameter Name	Type	Description
num_control	integer	The number of units in the control group
num_treated	integer	The number of units in the treated group
num_cov	integer	The number of covariates
bernoulli_param	numerical	The bernoulli parameter determining the distribution of the covariates
bi_mean	numerical	The treatment effect is normally distributed with this mean
bi_stdev	numerical	The treatment effect is normally distributed with this standard deviation

Return Name	Type	Description
data_frame	Pandas Dataframe	A dataframe that is according to the specifications. Units 0 through num_treated are treatment group, and num_treated through num_treated+num_control are control group units.
true_catt	array	The CATE of each treated unit, where the ith has the treatment effect of the ith unit in the dataframe.