A Friday AM thread on randomization inference (RI):

For a randomized experiment, a key difference between RI and standard statistical methods is that RI treats the sample as fixed and the treatment assignment as random, while standard statistics treats the sample as random In a simple t-test to compare a treatment to control, a p-value is obtained using the null distribution of the t-statistic. This null distribution comes from hypothetical repeats of this experiment (under H0), each with a different random sample from the population of interest.

Since my I got PhD in Biostatistics in 2020, I've worked on clinical trials design, as a postdoc, and now at as a pre-clinical statistican. A long🧵on my thoughts about the post-PhD job search, and why I don't like the academia vs industry dichotomy: I didn’t really start seriously thinking about what I wanted to do after the PhD until a couple of months before I defended. The end of the PhD is an exciting, but also very stressful time, and it’s very hard to make a huge life decision while also finishing your thesis.

The pre-print of our paper "Patient trajectories and risk factors for severe outcomes among persons hospitalized for COVID-19 in the Maryland/DC region" is finally out: medrxiv.org/content/10.110…. Really proud of the work done with @StrictlyStat @ScottZeger, and others. We show patient disease trajectories for all 832 patients in our cohort, classified by peak disease severity while in the hospital. We also show the percentage of patients in these disease states over time.