Repeated measures ANOVA - 2 groups and 3 time points - In need of Non para alternatives

#1
Hi, I've got skewed data and have done what I can to normalise the data with little success. I wanted to know in terms of Non-para alternatives what I could run within SPSS to get results for the data.
 

Karabiner

TS Contributor
#3
Therre is no "non-parametric" split-plot ANOVA.

ANOVA doesn't require a normally distributed dependent variable.

Prefereably, the dependent variable within each group (or: the residuals) should be sampled from a normally distributed population.

But even this is not necessary if n (total) is large enough. How large is your total sample?

With kind regards

Karabiner
 
#4
Unfortunately I've looked at the Studentized Residuals and they contain Outliers. The Normal Q-Q Plots are not normal either. I've attempted to transform the data using LOG10 but it makes the data much harder to interpret.

Sample wise I have 94 with 47 in each group.
 

Karabiner

TS Contributor
#5
I am not sure if it would be adequate to log transform the whole dataset because of some outliers, or whether ouliers are adequately dealt with using log transformation, or - as you mention yourself - how log transformed could be interpreted within the context of your research. Unfortunately I forgot to ask about the topic of your research and what the dependent variable actually does represent, since dealing with outliers is not a pure technical question, but depends strongly on content.

I suppose that with n=94 and equal groups, outliers do not affect assumptions (maybe someone wants to correct me here), but perhaps results. You could perform your split plot ANOVA with the untransformed data, and for checking robustness you could then repreat the analysis without extreme outliers. Or, if you have a substantial knowledge about the processes which produce the outliers, you could incorporate them into your model.

With kind regards

Karabiner
 
#6
Thanks for the response Karabiner. There are in fact 8 different measures each being tested using an ANOVA. Two interventions being compared that help personal problem solving.
  • Problem-related Distress - one outlier, studentized residual value of -3.01. Normally distributed, as assessed by Normal Q-Q Plot.
  • Depression - three outliers, studentized residual values of 3.34, 3.90, 3.06. Not normally distributed, as assessed by Normal Q-Q Plot.
  • Anxiety - six outliers, studentized residual values of 3.11, 3.18, 3.86, 3.86, 3.86, 4.21. Not normally distributed, as assessed by Normal Q-Q Plot.
  • Stress - two outliers, studentized residual values of 3.36, 3.51. Not normally distributed, as assessed by Normal Q-Q Plot.
  • DASS-21 - three outliers, studentized residual values of 3.14, 3.28, 3.62. Not normally distributed, as assessed by Normal Q-Q Plot.
  • Helpfulness - no outliers. Not normally distributed, as assessed by Normal Q-Q Plot.
  • Problem resolution - three outliers, studentized residual values of 3.20, 3.20. Not normally distributed, as assessed by Normal Q-Q Plot.
  • Use again - no outliers. Not normally distributed, as assessed by Normal Q-Q Plot.