0

I'm working on my thesis and I've entered the data of my survey in SPSS. It's a 2x2x2 between subjects experiment and I'm researching whether sustainability of the product (yes/no), sustainability of the package (yes/no) and beauty of the package (yes/no) have an influence on the product quality perceptions ( three way or two way interactions or whatever I find in the analysis).

I want to perform a Three-way Anova but the assumption of equal variances are not met. Since there's difference packages I'm not sure whether that really matters? Can anyone tell me if I should continue my analysis or not? And since there's only 2 conditions for my independent variables I can't check for the post hoc results where assumptions are not met (such as Tanhame).

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
Axelle Vdp
  • 1
  • Anova assumes that equal variances. So, I think, you cannot go forward the Anova route. Why can't you go for non-parametric tests like Man-Whitney U test which are bereft of any prior assumptions on the data? – honeybadger May 27 '18 at 08:07
  • @kasa, Mann-Whitney cannot address a factorial design. However, [an aligned ranks approach](http://depts.washington.edu/madlab/proj/art/) may be applicable. – Sal Mangiafico Nov 24 '18 at 13:50
  • @ OP, There are several approaches to deal with heteroscedastic errors. See [this answer on CV](https://stats.stackexchange.com/questions/91872/alternatives-to-one-way-anova-for-heteroskedastic-data/91881#91881) – Sal Mangiafico Nov 24 '18 at 14:06

0 Answers0