I'm comparing RPE (6-20) before and after an intervention, I've seen criticism of using parametric tests as RPE is ordinal data but it also suggested F tests are more robust and can be used with ordinal data. The RPEs I will be comparing are already averaged from 11 participants over 10 different measurements: 1st measure avg, 2nd measure avg... 10 measure avg, then compare these with the ten averaged post measures for a significant difference. I've been using t-tests for my ratio data, is it fair to class this data as interval and continue with t-tests, or is ordinal more accurate and do i need a non-parametric test.
The simple answer is that ordinal model should be more accurate as there are situations where the analysis of the average of the ordinal scores won't give the same results as a model that's more faithful to the structure of the data being modelled.
There are various options here but I'd stick with a parametric model that is ordinal if I had sufficient data. This explains the kind of model you might consider (but there are others).
https://psyarxiv.com/x8swp/
Here is a useful reference that is available here on RG:
Article LIkert scales, levels of measurement adn the “laws” of statistics
Harry Mahoney , you didn't explain what RPE is. When I Google it, the first hit suggests it's "rated perceived exertion". Is that what you mean?
- https://my.clevelandclinic.org/health/articles/17450-rated-perceived-exertion-rpe-scale
@Bruce Weaver yes it's rate of perceived exertion, apologies for the lack of clarity
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis