Do you have to use a parametric based ANOVA? Have you considered using the non-parametric variant of an ANOVA?
So I am trying to do an one-way ANOVA on some data. I was able to meet the assumption for homogeneity of variance (used levene's test), but not normality (used shapiro-wilks test). I plotted my data on a histogram and noticed all of the data are "fun"....majority of the numbers are e^1, while about 1/5 are e^15. Soooo basically I have no idea if transforming this data will even do any good? Should I continue on with an ANOVA? Thanks!