Testing two competing explanations for the association between pathogen load and ethnic diversity
pathogens limit the size of states (Cashdan, citing Crosby and McNeill)
high pathogen load causes people to avoid other groups, leading to limited dispersal and xenophobia (Thornhill)
Preliminary analysis presented at AAA (see Cashdan for pdf) supports the former, not the latter.
Each case is a society (n=186) from the standard x-cultural sample. Flat files, some missing values. Variables:
Ethnic diversity. a ratio variable (number of ethnic groups in a given area)
Many social variables: ordinal, coded from ethnographies.
Many environmental variables: both ratio (rainfall, etc), and ordinal (pathogen load, etc.).
ordinal data limit use of many standard methods (GLM). Multivariate ordinal regression methods would be useful for this analysis.
Galton's problem. Phylogenetically-correct statistics might be useful, although not as much a concern for this question as for a typical cultural trait.
Some of these data were analyzed earlier for a different question: why is ethnic diversity greater near the equator? The analysis used chiefly bivariate correlations because the indep. envir. variables were so intercorrelated, but someone with more multivariate savvy than I had could do a better job. I would support a re-analysis. I can send the paper.
Elizabeth Cashdan, Dept of Anthropology, University of Utah