Robust Regression in Genomic Data Analysis

This lecture covers the concept of robust regression in the context of genomic data analysis, focusing on downweighting observations with large residuals to improve estimation accuracy. Various loss and weight functions are discussed, including the Huber loss function. The lecture also delves into the use of M-estimators and the iterative reweighted least squares (IRLS) algorithm for robust regression estimation. Practical applications of robust regression in microarray data analysis, outlier detection, and quality assessment of gene expression measures are explored through examples and pseudo-images. Quality assessment metrics such as NUSE and RLE are explained, providing insights into chip quality and expression measurement error. The lecture concludes with a discussion on different quality assessment measures and their implementation in genomic data analysis tools.