A research lab from a well-known American University contracted us to analyze data from a pilot study they conducted on 77 mice of 2 genotypes and 3 age groups. The data included various cellular measurements between 1 and 5 cells per mouse. The first goal was to test for any differences in those cellular measurements between genotypes within an age group, and between any two age groups within a genotype. The second goal was to establish a relationship between power and sample size for detecting a certain significant difference in some of the cellular measurements while controlling a 5% type I error.

We performed descriptive analysis. For example, we made box plots of each cellular measure by genotype and age group, and computed the sample averages and the covariance structure. We ran a Generalized Estimating Equation (GEE) with an exchangeable correlation structure for each cellular measure using the geepack R package. For power and sample size calculation, we used and modified a method described in the paper “Sample Size Calculations for Studies with Correlated Observations” (Biometrics 53, 937-947, September 1997). The sample size and power relationship we calculated allowed the client to intelligently design a bigger trial with confidence.