What
Switch from log-ratio regression with Least Squares to multinomial logit regression
Why
Multinomial logit regression uses the correct statistical model for count dataacross categories, while OLS makes assumptions that are violated by this type of data (independence, normality, equal variances).
How (optional)
Deliverables
- Working, updated pipeline that produces interpretable tables on the correlations (or lack thereof) for the variables of interest.