Piffer’s method to examine the between group heritability of cognitive ability and height uses polygenic scores (either by simple mean or with factor analysis) based on GWAS findings to see if they predict phenotypes for populations. The prior studies (e.g. the recent one we co-authored) have relied on the 1000 Genomes and ALFRED public databases of genetic data. These datasets however do not have that high resolution, N’s = 26 and 50. These do not include fine-grained European populations. However, it is known that there is quite a bit of variation in cognitive ability within Europe. If a genetic model is true, then one should be able to see this when using genetic data. Thus, one should try to obtain frequency data for a large set of SNPs for more populations, and crucially, these populations must be linked with countries so that the large amount of country-level data can be used.
The above would be interesting and probably one could find some data to use. However, another idea is to just rely on the overall differences between European populations, e.g. as measured by Fst values. Overall genetic differentiation should be a useful proxy for genetic differentiation in the causal variants for cognitive ability especially within Europe. Furthermore, because k nearest spatial neighbor regression is local, it should be possible to use it on a dataset with Fst values for all populations, not just Europeans.
Since I have already written the R code to analyze data like this, I just need some Fst tables, so if you know of any such tables, please send me an email.
There is another table in this paper: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2675969/
The study is also interesting in that they note that the SNPs that distinguish Europeans the most are likely to be genic, that is, the SNPs are located within a gene. This is a sign of selection, not drift. See also the same finding in http://www.ncbi.nlm.nih.gov/pubmed/18246066/