Does phoneme inventory size correlate with speaker-population size once genealogy and area are controlled? A PHOIBLE-scale test

Hay & Bauer (2007) reported a positive correlation between a language's number of speakers and its phoneme inventory size; Moran, McCloy & Wright (2012) argued the effect is not robust under proper sampling and controls. QUESTION: using PHOIBLE inventories (segment counts per language/doculect) joined to open speaker-population estimates (e.g. Wikidata property P1098 via ISO-639-3/Glottocode), estimate the association between inventory size (total, and split into consonant and vowel counts) and $\log(\text{population})$, first uncontrolled and then controlling for genealogical and areal non-independence (a mixed model with family and macroarea, or a family-and-macroarea-stratified permutation test); report the sign, magnitude, and significance, and test robustness to PHOIBLE's multiple-doculect-per-language selection rule.

Acceptance. FULLY RESOLVES: join PHOIBLE 2.0 inventories to open speaker-population estimates for $\geq 300$ languages; report the uncontrolled correlation of inventory size with $\log$ population, then a controlled estimate (mixed model with family + macroarea, or stratified permutation) with coefficient sign, size, and CI, plus a robustness check across at least two PHOIBLE doculect-selection rules; ship the join + analysis script from the public CLDF/Wikidata data. PARTIAL: a faithful replication of Moran et al. (2012) on current PHOIBLE, or the uncontrolled full-sample correlation plus a family-level aggregate.