|
Abstract:
|
One can use factorial designs for sample size estimation in microarray studies. Sample size calculations follow the multiple testing framework given by Benjamini and Hochberg (1995), which limit the number of false positive and negative determinations of differential expression. The method for a 2X2 design follows: 1) set alpha and beta and calculate a sample size for a single gene analysis such that: among G independent genes (with G0 and G1 genes not differentially and differentially expressed, respectively), the probability that phat (the proportion of genes that test as differentially expressed among the tests showing differential expression) is greater than "p" (e.g., 0.95), will have a specified confidence (e.g., 99%) 2) divide the required sample size in half to obtain the number for each of the four groups. An example will be provided from the Lung Tissue Research Consortium.
|
