PCM3: Primate Continuous And Categorical Regression¶
PCM3 is the multivariable regression study in the evidence book. It stays with the primate dataset, but the main question changes: how should readers interpret models once continuous predictors, categorical predictors, interactions, and phylogenetic correction all appear together?
flowchart LR
A[Continuous predictor] --> B[Categorical predictor]
B --> C[Interaction term]
C --> D[Phylogenetic regression]
Why PCM3 Matters¶
PCM3 is where the evidence book stops sounding like "PGLS" as a single broad label and starts showing how interpretation really changes when coding, standardization, interaction terms, and phylogenetic correction all matter at once.
What PCM3 Helps You Read¶
- how the intercept changes meaning after standardization or transformation
- how categorical predictors depend on a clear coding choice
- how interactions should be read as joint effects rather than isolated slopes
- how the regression story changes again after phylogenetic correction
Checked-In Study Dossier¶
The full checked-in record lives under
evidence-book/studies/primate-continuous-and-categorical-regression/.
Inputs And Evidence Map¶
- lecture source copies:
primate_raw.xlsx,primate.RData, andprimatetree.nex - governed derivatives:
reference_primate.csvandreference_trimmed_primatetree.nwk - catalogs: dataset catalog and provenance guide
- evidence bundles:
evidence-001throughevidence-008
Read PCM3 In Order¶
What To Take From PCM3¶
- the repository now exposes a structured regression study rather than a single undifferentiated "PGLS" claim
- interaction and coding choices are part of the scientific interpretation, not merely model syntax
- deployed bundles improve reviewability even when lecture-specific verdicts remain narrower than PCM1 or PCM2
What Not To Overread¶
- bundle deployment is not a blanket claim of full numerical parity
- PCM3 does not replace the need to inspect the checked-in source record when provenance matters
- the study remains narrower than a full closed lecture-equivalent suite