I am examining whether sex and religion of a defendant may impact their percieved guilt, risk, possibility of rehabilitation and the harshness of sentencing. I have done this by creating 4 different case studies in which a defendant has differing sex and religion and was suspected of committing a crime. There were 200 participants in which 50 each where given one of the 4 case studies. Participants would then have to answer a number of questions about the case study such as "what sentence do you think is fair?" All data is ordinal. Ive been advised to used different statisical analysis so im confused and would like some advice on which one to use
If the data are ordinal then it violates assumptions underlying many types of tests (e.g., ANOVA). Thus, the non-parametric equivalent is an option. Kruskal-Wallis is potentially suited for this design. I've done comparisons of one-way and factorial anova versus their non-parametric equivalents and found the results to be virtually identical. If the DV is on a 7-point scale (or similar) then then it is technically ordinal but practically interval in nature, and most researchers use parametric tests. If it is truly ordinal, such as a rank-based DV, then you'd want to look at rank ordered tests (e.g., Friedman's test), but I doubt that would be a good fit here. I think you could also set up a regression model depending on how comfortable you are with it.
- Badges
- Science topic
- Similar topics
- Psychology