Hi, I'm doing several 2x2 ANOVAs as part of my MSc research project in psychology- two on my overall sample (n=69) looking at the effect of depression x stress on negative affect (NA) and positive affect (PA) overnight. Then I'm doing four on a sub-set who recalled a dream (n=47) looking at the effect of depression x dream valence and depression x dream intensity on NA and PA.

My DVs (NA and PA) are mean scores from 10-item Likert scales (the PANAS).

I've plotted the residuals for each set of ANOVA groups (splitting my SPSS file according to the ANOVA categories).

I get a couple of extreme outliers (3 SDs from box plot edge) and outliers (2 SDs from box plot edge) in each ANOVA.

I've looked at the outliers and I can't see any evidence that they were entered incorrectly or that participants didn't follow the procedure. I've also tried running the ANOVAs without the extreme outliers and it does change my result (Equality of Variance assumption is violated and one of my non-significant main effects becomes significant).

I've also tried running the data with the sum scores from the LIkert scale (rather than the mean). Confusingly, I then find DIFFERENT outliers.

I'm wondering if I should just include the outliers and report them in my analysis, suggesting that my results must be interpreted with caution.

What do others think?

Thanks so much,

Sarah

More Sarah O'Reilly's questions See All
Similar questions and discussions