A genome wide association study may have spurious or misleading results due to population stratification. This research evaluated the properties of global principal components and local principal components to adjust for population stratification. Principal components were calculated using both common variants (with minor allele frequency greater than 0.05) and rare variants (with minor allele frequency between 0.0005 and 0.05). One genetic model considered was from the Genetic Analysis Workshop 17 (GAW17). Additional genetic models developed in these analyses used the genotypes in the International Hapmap data. Phenotypes were simulated using these genotypes. Both type I error rates and powers of different models for identifying genetic variants associated with a phenotype were assessed. The four models in these analyses were: (1) using the number of minor alleles as the predictor variable for the phenotype; (2) using the number of minor alleles and 10 global principal components as the predictor variables for the phenotype; (3) using the number of minor alleles and 10 local principal components as the predictor variables for the phenotype; (4) using the number of minor alleles and the self-reported population of the participants as the predictor variables for the phenotype. Both the global PC adjustment model and local PC adjustment model had null hypothesis rejection rate roughly equal to the nominal significance level and comparable power to detect the causal genes. Both had better rejection rates than the model using the self-reported population indicators.