You could use engagement rate as a covariate in the analysis (ANCOVA is the typical way to do this). But really if you actually do the randomization correctly you really shouldn't end up with a case where one group has a significantly higher engagement rate on average.





Reply With Quote

