Robust Association Analysis Method for Count Trait Locus Mapping

In the era of big data, count data as an important type of discrete data has become a cutting-edge topic in statistical and biomedical research. Conventional association analysis approaches frequently prove inadequate when handling such data, urgently necessitating breakthroughs in new statistical methodologies. To address this crucial scientific problem, this study innovatively proposes a robust analytical method based on an inverse regression framework. The inverse regression model, as a novel methodological approach, ingeniously overcomes the limitations of traditional methods in handling ordinal response variables by treating traits as predictors and genetic markers as response variables. Since classical testing methods are sensitive to Hardy-Weinberg equilibrium (HWE) assumptions and struggle to account for covariate effects or sample correlations, their practical applications are constrained. Therefore, this paper develops RA-CT, a novel generalized robust method for locus-specific genetic mapping tailored for count data. First, we propose an inverse regression model using individual alleles as response variables, where the derived test statistic from this unified regression framework incorporates an innovative correction factor that explicitly adjusts for potential deviations from HWE. Second, by categorizing traits through indicator functions, the model accurately captures the distributional characteristics of count data. Finally, several tests are included as special cases, demonstrating the theoretical completeness of the approach. Simulation studies show that the new RA-CT method significantly outperforms existing approaches in controlling type I error rates and enhancing test power, providing a powerful analytical tool for genetic mapping of count traits.