Chapter 1 About

This is a book written in Markdown. Please contact Sarah Urbut for questions.

I’m really excited about these latent dirichlet allocation models, specifically for their flexibility in modeling time dependent changes in comorbidity profiles. However, I think much remains to be discovered. In the following pages, I hope you’ll enjoy:

1.1 Overview

First, we discuss the general framework for topic models, and introduce typical notations and probability distributions

Second, we describe the framework in McVean et al, particularly with reference to the use time dependent per disease trajectories

Third, we introduce two novel (and we hope clinically relevant) features:

1.2 Summary

Namely: while a topic may capture heterogeneity in diseases over time, it does not capture the fact that an individual’s topic distribtuion may change dramatically over time

Second, genetics can impact the speed through a topic, such that the time scale of the spline is changed dramatically.

We simulate these changes accordingly in the text that follows. So please, fasten your seatbelts and prepare to enjoy!