Application of mixed model ANOVA in microarray data analysis
by
Gary A. Churchill
The Jackson Laboratory
Coauthors: Xiangqin Cui, Natalie Blades

The correlation among observations in microarray experiments complicates the problem of testing for differential expression. We explore the application of mixed model ANOVA to microarray experiments with general design structure including replication at multiple levels within the experiment. By modeling some of the design factors as random effects, we can gain insights into the relative importance of different sources of variation. Estimated variance components provide a basis for allocating replication to different levels of the experiment. They also provide guidance for technology improvement by pin-pointing the largest sources of variation. Furthermore, they can provide interesting insights into the nature of intrinsic biological variation of gene expression.

Date received: September 4, 2003