Check out our new preprint on a new parametric approach to survival analysis https://arxiv.org/abs/2009.10264  #epitwitter with #rstats package https://cran.r-project.org/package=casebase (THREAD 1/n)

Did you know that Cox himself isn't a big fan of the Cox model? (2/n)

The casebase approach allows you to fit smooth-in-time hazard functions which means that you get nice interpretable curves for the cumulative incidence instead of step functions (3/n)

By directly modeling time, we can also readily obtain non-linear time-dependent hazard ratios (4/n)

Our approach can also handle competing risks data, time-varying exposures, and perform variable selection (5/n)

We leverage the key insight made by Hanley and Miettinen (2009) that continuous-time survival modeling can be reformulated as a logistic (or multinomial) regression with an offset term (6/n)

In a sea of survival analysis packages, which one should you choose? We provide a comprehensive survey of #rstats packages and summarize their key features (7/n)

We also provide a new exploratory analysis tool for visualizing survival data in the form of population-time plots (8/n)

This work was jointly co-authored with current Assistant Professor of Statistics and Computer Science at UManitoba @mturg1989. We started this project while doing our PhD in the same lab. Here is a picture of our outline taken in our shared office August 19, 2015 (9/n)

Admittedly, we didn't work on this project continuously for the past 5 years. Nevertheless, it's an example of how long it can take to bring a project to completion. I'm proud that we finally got this done with the help of my star PhD student @jesse_islam. (end of thread)