opossum 0.2.1

Simulated Data Generating Process

Opossum

Package for simulation of data generating process to evaluate causal inference models.

Getting Started

Latest release version: 0.2.0

Download with pip

*nix OS

pip3 install opossum

Windows

pip install opossum --user --upgrade

Build it yourself

git clone https://github.com/jgitr/opossum.git
cd opossum
git checkout master
cd opossum

*nix

python3 main.py

Windows

python main.py

SHA-256: opossum-0.2.0-py3-none-any.whl

19b0b705b37a71fd5deac40720d30da4caa6529a9f1e1e7db648ecd07aea3077

SHA-256: opossum-0.2.0.tar.gz

e13000b1755576a80693dca5f2560cc9b7d3e1391701bdc5cae86f3a9b7eb036

Application

Bellow you can find a short description of the core functions that the package offers in code form. For more detailed information on how to apply the package and to get insight into the theoretical model that it is based on, please refer to the following blog post https://humboldt-wi.github.io/blog/research/applied_predictive_modeling_19/data_generating_process_blogpost/.

Default Setting

from opossum import UserInterface
# number of observations N and number of covariates k
N = 10000
k = 50
# initilizing class
u = UserInterface(N, k, seed=None, categorical_covariates = None)
# assign treatment and generate treatment effect inside of class object
u.generate_treatment(random_assignment = True, 
                     assignment_prob = 0.5, 
                     constant_pos = True, 
                     constant_neg = False,
                     heterogeneous_pos = False, 
                     heterogeneous_neg = False, 
                     no_treatment = False, 
                     discrete_heterogeneous = False,
                     treatment_option_weights = None, 
                     intensity = 5)
# generate output variable y and return all 4 variables
y, X, assignment, treatment = u.output_data(binary=False, x_y_relation = 'partial_nonlinear_simple')

Choosing covariates

N = 1000
k = 20
# whole dataset is binary
u = UserInterface(N, k, categorical_covariates = 2)
# one quarter of the dataset is binary
u = UserInterface(N, k, categorical_covariates = [5,2])
# dataset consists of 10 continuous, 4 binary, and 3 variables each with 3 and 5 categories respectively 
u = UserInterface(N, k, categorical_covariates = [10,[2,3,5])

Creating treatment effects

# random treatment assignment resulting in on average 20% treated observations 
u.generate_treatment(random_assignment = True, assignment_prob = 0.2)
# non-random treatment assignment with on average 65% treated observations
u.generate_treatment(random_assignment = False, assignment_prob = 'high')
# generating only a positive heterogeneous treatment effect
u.generate_treatment(constant_pos = False, constant_neg = False, heterogeneous_pos = True, heterogeneous_neg = False, 
                     no_treatment = False, discrete_heterogeneous = False)
# generating a heterogeneous treatment effect that is in 20% of cases negative and 80% positive
u.generate_treatment(treatment_option_weights = [0, 0, 0.8, 0.2, 0, 0]) 

Creating output

# Creating continuous y with partial nonlinear relation 
y, X, assignment, treatment = u.output_data(binary=False, x_y_relation = 'partial_nonlinear_simple')
# Creating binary y with underlying linear relation and added interaction terms of X
y, X, assignment, treatment = u.output_data(binary=True, x_y_relation = 'linear_interaction')