Hello everyone,

Let's say I'm gonna compare three groups of something and that the distribution of these groups does not meet normality nor homoscedasticity (mr. Shapiro-Wilk and Mr. Levene's told me so).

I think I might try to transform such data (SQRT, LOG10...) and see if transformed data met the assumptions. How should I transform the data given that one of the three groups shows a negative skewness? (see the picture attached)

For example, I try first with SQRT-transformation and thus I compute =SQRT(y) in Excel. This should work for positively skewed data but for the negative ones. In that case, I think I should reflect the formula using =SQRT(max value+1-(y)).

So, basically my question is: Is it correct to apply the reflect transformation only for the negative skewed group (Group C, colored in yellow in the attached figure)? If not, how to deal with negative- and positive-skewed groups?

Thank you!

More Matteo Francioni's questions See All
Similar questions and discussions