Comparison between frequentist and Bayesian implementation of mixed linear model for analysis of microarray data
MetadataShow full item record
The objective of this study was to evaluate the performances of a mixed linear model under a frequentist and a Bayesian implementation for analysis of microarray data. A simulation was conducted following the structure of an existing Affymetrix chip data. PROC MIXED of SAS was used for the frequentist implementation. T-test, p-values, and the estimated difference between the two treatment levels were used to detect differentially expressed genes, as well as false positive and false negative cases. In the Bayesian implementation, the probabilities of a gene being in each of five pre-defined significance level classes were used for performances testing. The results indicate that both methods performed exceptionally well in identifying highly differentially expressed genes with a success rate of 0.96 and 0.98, respectively. However, the Bayesian approach was superior in clustering the most important genes. Both procedures performed similarity in detecting false positive and negative cases.