2018/06/11 by Olga A. Vsevolozhskaya, Vsevolozhskaya, Olga A, Dmitri V. Zaykin +1
Biochemistry, Genetics and Molecular Biology · Mathematics · #Advanced Causal Inference Techniques #FOS: Computer and information sciences #Gene expression and cancer classification #Genetic Associations and Epidemiology #Methodology (stat.ME) #Statistical Methods and Bayesian Inference
paper · pdf · doi:10.48550/arxiv.1806.04251
openalex publication_date 2018/06/11 · openalex created_date 2022/09/14 · openalex updated_date 2026/07/28
The odds ratio (OR) is a measure of effect size commonly used in\nobservational research. OR reflects statistical association between a binary\noutcome, such as the presence of a health condition, and a binary predictor,\nsuch as an exposure to a pollutant. Statistical inference and interval\nestimation for OR are often performed on the logarithmic scale, due to\nasymptotic convergence of log(OR) to a normal distribution. Here, we propose a\nnew normalized measure of effect size, \γ', and derive its asymptotic\ndistribution. We show that the new statistic, based on the \γ'\ndistribution, is more powerful than the traditional one for testing the\nhypothesis H0: log(OR)=0. The new normalized effect size is termed `gamma\nprime' in the spirit of D', a normalized measure of genetic linkage\ndisequilibrium, which ranges from -1 to 1 for a pair of genetic loci. The\nnormalization constant for \γ' is based on the maximum range of the\nstandardized effect size, for which we establish a peculiar connection to the\nLaplace Limit Constant. Furthermore, while standardized effects are of little\nvalue on their own, we propose a powerful application, in which standardized\neffects are employed as an intermediate step in an approximate, yet accurate\nposterior inference for raw effect size measures, such as log(OR) and\n\γ'.\n