American Statistical Association New York City Metropolitan Area Chapter Memorial Sloan-Kettering Cancer Center Biostatistics Seminar

Assessment of the Hardy-Weinberg proportion (HWP) in controls has been widely used as a quality control measure in case-control association studies. However, when the disease being studied is common, controls might not represent the general population, which could result in inaccurate HWP test results. Such results could lead investigators to discard important single-nucleotide polymorphisms (SNPs) that could potentially be causal. In this talk, I will show the inappropriateness of the HWP test in controls and propose a mixture HWP (mHWP) exact test using a mixture sample that mimics the general population. The simulation results showed that the mHWP exact test is more powerful than either the traditional HWP method in controls or the likelihood-based approach for keeping causal SNPs for further analysis when the disease is more common. The approach also can maintain good control over type I error, even when genotyping errors exist.