The repository is for the paper:
Description of the files here (please read the paper for understanding the references):
-
Datasets:
a. strong_bias_data.csv (simulated dataset with strong bias through confounder)
b. weak_bias_data.csv (simulated dataset with weak bias through confounder)
c. ewing-data.csv (real-world dataset)
d. adj_surv_data_transformA.csv (adjusted datasets through SCM, using transformation A)
e. adj_surv_data_transformB.csv (adjusted datasets through SCM, using transformation B)
-
Main code:
a. 1_simulate-data.ipynb (code for simulation of data)
b. 2a_ipw_simulated_data.ipynb (code for adjustment on simulated data)
c. 2b_ipw_ewing_data.ipynb (code for adjustment on real-world data)
d. ipw_ewing_data_new_graph.ipynb (code for adjustment on real-world data, using transformation B)
-
Images:
a. unadjusted.eps & unadjusted_ewing.eps (unadjusted/direct survival curves for both simulated and real data)
b. adjusted.eps & adjusted_ewing.eps (adjusted survival curves for both simulated and real data)
-
Extras:
a. or_rr_simpsons_paradox.ipynb (code for similar explorations on odds and risk ratio, already known in literature as Causal Odds Ratio and Causal Risk Ratio)
b. desmos.png (baseline for simulated data survival curve, plotted as mathematical functions)