The purpose of this research is to study statistical genetics problems in the area of mapping human trait variation. We have examined the multiple testing performance of single-locus and multi-locus mapping techniques such as the analysis of association of haplotypes with disease susceptibility and adverse reactions to drugs. Statistical properties of a common data mining technique (recursive partitioning) have been investigated. We studied chances of discovering genuine associations by employing statistical distributions of ranks of true associations. We considered multiplicity issues in the context of large scale mapping experiments, such as whole genome association scans. Our design incorporated dependencies between test statistics evaluated at different genetic locations due to linkage disequilibrium and haplotype blocks.
