ABSTRACT
This chapter presents basic statistical principles needed during analysis of array data, in the hopes that scientists will be able to make better decisions with their data. Microarray experiments certainly illustrate the definition, with scientists trying to decide which genes show differential expression in the face of substantial biological and technical variation. Hypothesis testing for comparing means is commonly conducted by t-tests or analysis of variance, appropriate when the measured response variable is normally distributed. The log transformation generally improves normality, but also usually stabilizes the variance, or makes error variances more equal. The design of experiments using oligonucleotide arrays differs from cDNA arrays in that only a single set of treatment conditions is applied to an array, rather than having two treatment conditions applied as with the cDNA array. Chu et al. recommend a mixed-models approach to the analysis of oligonucleotide arrays.
