Mixed Anova (SPSS) data transformation irregularities - inaccurate results

#1
Hi! :)

I am evaluating an intervention program aiming to increase the Stress Management skills of learners. To evaluate the program's impact I am doing a quasi-experimental study with a treatment group and a control group. Both groups were assessed at baseline and again after the program ended (one year).

Means increased slightly from baseline to post testing:
• Baseline Treatment Group mean: 37.41
• Baseline Control Group mean: 34.26
• Post Treatment Group mean: 37.51
• Post Control Group mean: 35.91

When checking assumptions I found that the distribution (for each group at each time point) was still slightly negatively skewed even after outliers were removed. Shapiro-Wilk was p < .05.

To convert the moderately negatively skewed baseline dependent variable to normality, I first attempted to apply a "reflect and square root" transformation. Basically, I found the largest Stress Management score (which was 47) in the baseline sample and then added 1 to this value. Then I created a new variable so that each score is subtracted from this value and then the square root of the scores taken. SQRT(47 – Stress Management)". I did precisely the same for the post Stress Management variable.

All distributions became normally distributed (Shapiro-Wilk: p > .05).

Now what is really strange is that when I run the Mixed Anova with these new transformed DVs, treatment and control group means decrease after a year.

Baseline Treatment Group mean: 3.19
Baseline Control Group mean: 3.66
Post Treatment Group mean: 3.11
Post Control Group mean: 3.38

If I run the Mixed ANOVA without transformed data, the means increase from baseline to post testing. How is this possible? I know the learners Stress Management abilities increased slightly over the one year period (as the original group means increased), but the transformed data says it decreased.

Perhaps I am doing something that is obviously wrong, but I can’t figure it out. I have run this analysis a number of times and double checked that I enter the right data at the right timepoints. I have done the same thing with an Intrapersonal Skills variable and the same happened. What am I doing incorrectly? If anyone can please give me some advice or explain what is happening?

Thanks for the help in advance!
 
Last edited: