# Mixed ANOVA - normal distribution with heterogeneity - transformation advice

#### KenSoona

##### New Member
Hello,

First time poster here. Any advice is appreciated:

I'm doing analysis of clinical trial data using SPSS. I'm doing a mixed ANOVA looking at score differences at three points in time (baseline, middle, final) for two groups of people (Intervention, Control). N=76, intervention n=26, control n=50

Using Shapiro test of normality, my data are normally distributed at each time point for each of the two groups, although the sig is below 0.10 for the control group at the baseline and final.

Using Levene test, I have heterogeneity of error variance between intervention and control groups, for the baseline and middle datapoints.

Sphericity is just below 0.05 ... 0.48

Looking at Skew and Kurtosis by group, intervention is skew = -0.246, kurt is -0.918, control is skew = -0.503 and kurt is -0.431, IE negative skew and flat values.

So...normal distribution, heterogeneous error variance, non-sphericitous, negative skew with flat values.

My question is how to transform the data, and should I even do it. I assume the first and second answers can be grouped as "yes, if you want your results to be accurate," but I have never transformed data before, so I'm going into semi-uncharted territory.

I apologize if this is covered elsewhere in depth on the site. I searched around and did looked around on google, but wasn't able to find an example of a similar situation that included how the analyst addressed the problem.

Any help would be very much appreciated!!