construct a multuplet with 3 cell types (for example) then while iteratively reducing the amount of one cell type, increase the amount of the other 2 running optimisation each time. What percentage is the lowest level that the optimisation can detect?
you can also try reducing the number of samples in each cell type. As the groupMean becomes less representative of the "true" mean, the optimisation should struggle
construct a multuplet with 3 cell types (for example) then while iteratively reducing the amount of one cell type, increase the amount of the other 2 running optimisation each time. What percentage is the lowest level that the optimisation can detect?