|
Abstract:
|
Precisely controlled cell line experiments usually use very few replicates. Coefficients of variation are typically under 30% and effects are large justifying the use of just 3 replicates. The small number of replicates leaves few options for the statistician for calculating p-values. In a project, several distinct experiments using the same cell lines are run independently. Although each experiment asks a different question, the measures will typically be the same. We propose taking advantage of the several experiments to "borrow" information from each other when determining error variances. The approaches we propose are similar to smoothing error variances in micro-array studies. Cell line experiments are distinct from micro-array studies, because of the smaller number of replicates, well-defined hypotheses, and fewer distinct experiments for smoothing variances.
|
