I plan to run analysis of variance with a data set. When I plot the residuals of YIX, it shows two distinct clouds of data. However, the histogram of the same residuals fits fairly to a bell shaped curve (see attachment).
Classic transformation approaches, such as log, quadratic, inverse or square root don't improve the shape of the plot of residuals.
It is a case of bi-modal data, which is explained by the nature of the experiment.
My question is whether anyone could assist me tackling this issue.
Another question is how trusting ANOVA results wiould be if I decided to use 'as-is' data anyways?
Thank you in advance for any help!
Luciano