This repository contains the (ugly research grade) code and data used for my paper on treating reproducible research as a survival analysis problem. The idea is that we can quantify (in a more objective but still not perfect) the nature of replication attempts by how much time was spent, rather than a binary "yes/no" answer. This way we do not preclude success, and avoid false claims of non reproduction for papers that require significant effort, or in cases where the reproducer's background is not well suited to the paper at hand (e.g., I'm just not smart enough to replicate in areas I've never worked in).
This work is a continuation of my prior research here, and was extended by looking up completion dates in github for as many algorithms as possible, and comparing with start dates in my Mendeley to get an estimate of time spent replicating.
In the future I hope we will all start recording time we spend trying to implement algorithm, as it will be valuable information for future studies on reproducibility.
If you use this work, data, or code, a citation would be appreciated!
@inproceedings{Reproducibility_Survival,
author = {Raff, Edward},
booktitle = {The Thirty-Fifth AAAI Conference on Artificial Intelligence},
title = {{Research Reproducibility as a Survival Analysis}},
year = {2021}
}