Recent Writing

Translating Causal Notation

I’m often frustrated by the myriad of hidden assumptions in so many “causal inference” methods, for example all of the structural assumptions that have to be made in order to turn a so-called causal graph into a fully-specified statistical model.  

For my supporters on patreon dot com a few months ago I dug into the assumptions implied in the “Table 2 Fallacy” paper of Westreich and Greenland, showing how those assumptions can be made explicit by embracing a probabilistic modeling perspective, https://www.patreon.com/posts/110488559.

Along the way I translated their causal graphs into corresponding variate-covariate models (https://betanalpha.github.io/assets/case_studies/variate_covariate_modeling.html), only without forcing specific linear regression structural assumptions. In my option this makes the role of the infamous “U” variable much less mysterious.

Mixture Modeling

Trying to decide between two data generating processes? Why not use both with mixture models?

HTML: https://betanalpha.github.io/assets/chapters_html/mixture_modeling.html

PDF: https://betanalpha.github.io/assets/chapters_pdf/mixture_modeling.pdf

In this chapter I cover the foundations of probabilistic mixture modeling, from the marginalization of categorical variables to numerically stable implementations and the specter of inferential degeneracies. Along the way I also discuss the special case of discrete and continuous inflation models.

Useful Reference

Some gems of the older internet: pretty comprehensive etymology research for some of the most ubiquitous terms in contemporary statistics, https://web.universiteitleiden.nl/fsw/verduin/stathist/1stword.htm and https://mathshistory.st-andrews.ac.uk/Miller/mathword/.

Consulting and Training

If you are interested in consulting or training engagements or even commissioning me to create a presentation on a topic of interest then don’t hesitate to reach out to me at inquiries@symplectomorphic.com.

Probabilistic Modeling Discord

I have a Discord server dedicated to discussion about (narratively) generatively modeling of all kinds, https://discord.gg/QKmrk4hy.

Support Me on Patreon

If you would like to support my writing then consider becoming a patron, https://www.patreon.com/betanalpha. Right now my covector+ supporters are enjoying early access to my next big chapter on pairwise comparison modeling techniques.