This repository contains replications of various publications using the zEpid package. When I find a publication (with publicly available data) that uses a method implemented in zEpid, I will attempt to replicate the corresponding analysis.
The purpose of the replication process is; (1) verify functions within zEpid, (2) demonstrate the utility of zEpid, and (3) go through use-cases of zEpid as a way to user-test the library
Current studies replicated are;
- Keil AP, Edwards JK, Richardson DR, Naimi AI, Cole SR. The parametric G-formula for time-to-event data: towards intuition with a worked example. Epidemiology (Cambridge, Mass). 2014;25(6):889-897. doi:10.1097/EDE.0000000000000160.
- Cole SR, Hernan MA. Adjusted survival curves with inverse probability weights. Comput Methods Programs Biomed. 2004;75(1):45-9.
Additionally, tutorials (which will be later added to the website) are available here.
If you have any recommendations on studies you would like to see replicated with zEpid, you can request either through GitHub or gmail (zepidpy)
The following code will only run with zEpid version 0.3.0+. In this version, both data sets are available to load directly
The following are replications of publications that use the g-formula
Using a dataset of bone marrow transplant recepients (n = 137), the marginal effect of preventing graph-versus-host disease compared to the natural course. SAS code for the implementation is available at https://github.com/alexpkeil1/Gformula-tutorial
zEpid-replications/Keil_2014 contains:
gformula_keil.ipynb (code and results)
The following are replications of publications that use IPW (any type)
Cole and Hernan demonstrate adjusting survival curves via inverse probability of treatment weights (IPTW) using Ewing's sarcoma dataset (n = 76).
zEpid-replications/Cole_2014 contains:
ipw_cole.ipynb (code and results)
Example of inverse probability of missing weights for a single missing variable. The example uses the sample data zet included with zEpid
Tutorials/IPMW_single.ipynb