normalising data for a mixed anova

I am running a 3x5 mixed factorial anova. I am testing the effects of 3 training groups over 5 time periods on the same questionnire.
The groups means at test 1 vary a lot and I have been advised to normalise the data by using z scores.
My question is do I then run the same mixed anova anaylsis using the z scores as I was doing with the raw scores and if so how do I then interpret the effects/interaction?