Use t tests to analyze counts of cells in two states

Some experiments generate data consisting of counts of cells in two states, such as alive or dead, CD3⁺ or CD3^-, etc. These are often converted to percentages and analyzed with parametric statistical tests intended for means like t tests, which assume data are normally distributed around means. This may not be true when percentages are close to either 0 or 100. I compared the performance of a comprehensive set of statistical tests intended for count data to t tests under conditions relevant to common wet-lab experiments, 3-6 replicate with 100-1000 cells per replicate. Only a few did as good a job as t tests at maintaining the rate of false positives, but those offered no advantage at finding differences when they exist. Researchers can use t tests to analyze count data converted to percentages under typical wet-lab conditions confident that this is an appropriate and effective approach.