Refactoring to Log-Ratio Regression

[griffinsharps]

What

This changes the regreession in the is analysis from a Multinomial Logit to a Log-Ratio Regression.
The analysis now aims to answer the question:
"How do block group demographics affect the distribution of clusters?"

Why

Previous version of the code answered the question:
"Given a household's demographics, what cluster is it in?"
Which was not the aim of this analysis.

How

1. Change from long format (multiple rows per block group) to wide format (one row per block group)
2. Replaced single MNLogit model with separate OLS/WLS per non-baseline cluster
3. Change coefficients from "odds ratios" to "multiplicative effects on proportion ratios"

Deliverables

• Summary statistics and basic findings for August 2023

[linear]

SMA-33