Department of Epidemiology and Biostatistics, School of Public Health
University of Illinois at Chicago
“Estimation of Strength of Signals Detected by Multivariate Test with Application to Circadian Rhythm”
In detecting genes controlling circadian rhythm, a chi-square test with two degrees of freedom is often employed. Since signals of circadian rhythm in gene-expression may not be strong enough to be detected at a high significance level required for adjusting for multiple tests, it is of substantial interest to estimate the total signal strength detected at a relatively low significance level. We propose an approximately unbiased estimator for the total detected signal strength at a given significance level. An inference procedure for the detected signal strength is also obtained. Simulation study demonstrates good performance of the proposed approach. The proposed approach is applied to the estimation of the total detected signal strength of circadian rhythm genes expressed in human brain tissue.