I've got a binary variable : 0 = they used something; 1 = they did not, then I have a total score out of 8 trials for each child and I originally conducted an ANOVA on the scores (that I actually transformed in percentages of use).
A reviewer is right in telling me that because of the original binary coding, I do not respect the normal distribution condition and I've got to apply an arcsine transformation to my data.
Any idea how to do it ?