Association studies offer great promise in dissecting the genetic basis of human complex diseases. For population based genetic case-control studies, perhaps the most commonly used procedure is to test genotype-phenotype association at each single SNP. It is widely believed that genotypic contributions to disease risks are generally no-overdominant, which means the heterozygote risk is intermediate between the two homozygote risks. Thus, it is possible to construct more powerful statistical procedure by using statistical tests tailored for this ordered restriction. In this chapter, we examined the statistical power and type I error rates of different statistical tests that are commonly used in single-locus association analysis. Our results indicated that although less powerful than allelic tests (i.e. 1-df Pearson X 2 or trend test) for near additive risk, the genotype-based tests (2-df Pearson X 2 or Fisher exact test) are generally more robust and powerful especially for risks far from additive and the power of genotype-based tests can be uniformly improved by applying the ordered restriction on genotypic risks.