Exploring the tradeoff between phenotypic specificity and sample size

In psychiatric genetics with genome-wide data we are forced to make a trade-off between well characterized phenotypes and large sample sizes. Well characterized phenotypes have less measurement error, and accordingly provide... [ view full abstract ]

In psychiatric genetics with genome-wide data we are forced to make a trade-off between well characterized phenotypes and large sample sizes. Well characterized phenotypes have less measurement error, and accordingly provide a higher signal to noise ratio in genomic analyses.  Due to the effort required to assess a well characterized phenotype, fewer observations can be collected for the same amount of effort/money. By contrast, if we relax the phenotypic specificity we can much more economically collect data, but the signal to noise ratio will be lower. Relaxing phenotypic specificity induces phenotypic heterogeneity. The ultimate question, therefore, is what is the optimal tradeoff in terms of statistical power between well characterized phenotypes and large sample sizes? Embedded within this research question is a series of hypotheses: a) what factors increase the statistical power in heterogeneous phenotypes and how much heterogeneity is tolerable within genome-wide association studies (GWAS), b) how much additional power is provided by well characterized phenotypes (e.g. where is the tipping point viz. lower sample size), and c) what meta-analytical or multivariate analytical? factors can be leveraged to enhance the power to detect significant genetic associations with varying levels of phenotypic specificity. These hypotheses will be tested using a series of simulation studies.