Dilsher Singh Dhillon

Last updated on Feb 2, 2020 12 min read

Hierarchical Multivariate Logistic Regression

Introduction Problem Set-up Data Hierarchical Model Prior Predictive Simulation Posterior Distribution Shrinkage Conclusion Introduction The original motivation for this post was to recreate an analysis in BDA (chapter 8), but as I started working through it, I realized it is an opportunity to write about several interesting concepts within the same framework. The overarching theme of this post covers 1. How to set up a multivariate logistic regression (not a multinomial or multivariable)

Last updated on Feb 2, 2020 12 min read

Multiple Comparisons and Bayesian Inference

Motivation Background Bayesian Inference and ANOVA Simulation Set up Frequentist pairwise comparisons Naive Tukey adjusted Multilevel Model Conclusion Motivation They say the best way to learn something, is to teach it! And that’s exactly what I intend to do. Inspired by Solomon Kurz’s blog posts on power calcualtions in bayesian inference, and Dr. Gelman’s blogs, here’s an attempt to open the guts of a problem I’ve struggled with.

Last updated on Feb 2, 2020 8 min read

Using tidy verbs in data.table

Introduction Results Code for simulation Conclusion Hello 2020! I’m going to start off this new year with a post that I’ve been meaning to put up for a while. It has nothing to do with statistics, but very much to do with my other love, the language R and it’s data manipulation libraries tidyr and data.table. Introduction For most of my work, the tidyverse set of packages get me where I want to be with the added ease of syntax, readibility and all the good stuff.

Last updated on Jan 1, 2020 7 min read

Estimating smooth trends and identifying periods of change using bayesian inference

Introduction Simulate data Model Estimating trends Estimating Derivatives Conclusion - Introduction As the title suggests, this post explores the powerful generalized additive models to model time-series data. More often than not, we associate modeling time-series data with forecasting i.e, understanding the underlying data generating process to then forecast the future. However, we could also be interested in understanding the nature of the current time-series, and also, comparing different time series.

Last updated on Feb 4, 2019 4 min read

How to git (hub)

I think everyone at somepoint, regardless of what industry they’re in, have folders which have files saved as _v2,_v3… v_final… That can only mean one thing down the line - chaos. I always though version control when it came to code was for a group of people working on a project, and for a lone statistician in the group as myself, I wouldn’t need one. But, Jenny bryan’s Excuse me, do you have a moment to talk about version control changed my mind.