A scratchpad for statistics, politics, and other odd lots
At best, Shapley values are misleading. At worst, they’re actively harmful. Don’t use them.
Oops I’ve been obsessed with sufficiency lately
One of the many ways to parameterize a sufficient Gamma
A reparameterization of the multivariate student-t distribution in Stan
Part 4: Adjustments for Aggregators
Part 3: Binary Surprises
Part 2: A Free Lunch
Writing this down to keep future me sane
An introduction to and description of a March Madness forecasting system
Part 1: The Abstract
What went well, what could be improved, and where to go from here
A power ranking for the title of most pedantic nerd on Dropout’s Um, Actually
The methodology behind the model
If you want to see a Democrat in the white house next year, Biden is your best bet
A survivor’s guide to estimating customer lifetime from discrete censored churn data
How to think about (and avoid thinking about) the presidential election in 500 days
A new metric for measuring regional partisanship
Reparameterizing multinomial models for better computational efficiency
Using gaussian process models for hospital goal setting
Estimating population results from non-representative survey samples using multilevel regression and post-stratification
Mixed scales lead to mixed results
Ordered categorical models for estimating Net Promoter Score: a hierarchical Bayesian implementation
A causal look at how Musk’s twitter ownership has affected Tesla stock
Multilevel Models for Patient Satisfaction
A look under the hood of the 2022 Midterm Model
A grammar of (nested) data manipulation
Modeling patient satisfaction data with empirical Bayesian methods
Improving a plot, one step at a time
Organizational resistance to uncertainty
How republicans who voted to impeach or acquit Donald Trump differ in their votes on democracy
Generate bootstrap prediction intervals from a tidymodel workflow!
Why you should put a roof on your house before hiring an interior decorator
Using bootstrap resamples to generate confidence intervals from non-linear models
Building a Presidential Approval Model
Building a Generic Ballot Poll Aggregator
Feature Engineering with the Diamonds Dataset
