While scCODA performs better than other methods in the low-sample case, we stress that analyses on datasets with larger sample sizes will always be less sensitive to outliers and variability in the data. To determine a reliable sample size for detecting effects of different strengths, we conducted a power analysis of our method.
