MATHEMATICAL PROBLEMS IN ENGINEERING, 2012 (SCI-Expanded)
The frequentist gene expression index (FGX) was recently developed to measure expression on Affymetrix oligonucleotide DNA arrays. In this study, we extend FGX to cover nonnormal log expressions, specifically long-tailed symmetric densities and call our new index as robust gene expression index (RGX). In estimation, we implement the modified maximum likelihood method to unravel the elusive solutions of likelihood equations and utilize the Fisher information matrix for covariance terms. From the analysis via the bench-mark datasets and simulated data, it is shown that RGX has promising results and mostly outperforms FGX in terms of relative efficiency of the estimated signals, in particular, when the data are nonnormal.