Miscellaneous statistical stuff

Current dysgenic estimates predict that genotypic IQ in the West are falling at a substantial rate, amounting to around half a standard deviation or more over the past century, by 1. reducing the frequency at which intelligence-increasing genetic variants occur (through natural selection against such variants) and 2. by increasing the number of new and potentially harmful genetic mutations (increasing mutation load). Estimates are produced indirectly by surveying reproductive rates or by trying to show decreases in phenotypic traits associated with intelligence; it would obviously be preferable to examine dysgenic effects directly, by observing decreases in frequencies or increases in mutation load in a large sample of Western genetic information such as SNP arrays or whole-genomes (respectively). Such direct testing of dysgenics hypotheses are becoming increasingly feasible due to the exponential decrease in SNP & whole-genome sequencing costs creating large datasets (some publicly available) and the recent identification of some intelligence genes. It remains unclear how large these datasets must be to overcome sampling error and yield informative estimates of changes in frequencies or mutation load, however; datasets like PGP or SSGAC may still be too small to investigate dysgenics. I considered the effect size estimates and under some simple models derive power calculations & power simulations of how large a dataset would be required to have an 80% chance of detecting a dysgenic effect: to detect the decrease in intelligence SNPs using SNP data, n≥30,000; to detect the increase in mutation load in whole genomes, n≥160. I then compare to available datasets: the effect on SNPs can be detected by a large number of existing proprietary databases, but there are no public databases which will be large enough in the foreseeable future; the effect on mutation load, on the other hand, can be detected using solely the currently publicly available dataset from PGP. So I conclude that while only the proprietary databases can directly test dysgenic theories of selection for the foreseeable future, there is an opportunity to analyze PGP genomes to directly test the dysgenic theory of mutation load.