Association mapping has been conducted in many different crop species for various complex traits. Besides the population structure issue that is inevitable in studying a diverse collection of germplasm, researchers often wonders about the sample size, number of backgound markers required, and their impact on statistical power. Our objectives are to: 1) determine the power of association mapping with different sample sizes; 2) determine the robustness of relatedness quantification with different numbers of markers and its effect on heritability estimation and statistical power. Empirical molecular and phenotypic data are used in computer simulation.