Hello, I am currently writing my masters dissertation based on mock jury perceptions and how they change based on a) the length of training a forensic expert has received at their job and b) the country they have gotten their degree from (whether it's local to the place the trial is happening or international). I by no means am great at statistics and haven't really done statistical work since high school. I'm a little confused about whether I should be using parametric or non-parametric tests.
I started by testing for normality, and out of the 16 different test results (I have 8 scenarios and tested each with a Kolmogorov-smirnov and Shapiro-wilk test), 13 yielded a normal distribution of over .05. So I assumed I should then use an ANOVA test to see if there's significance. However, when I double-checked this with a professor in my department, he (who said he also is no stats expert) said I shouldn't assume normality with any of my results, and stick to using a Kruskal-Wallis test over ANOVA. I'm not really sure why, but I'm wondering if someone could let me know what they think? My sample is 96 people if that has any importance. Thank you!
I started by testing for normality, and out of the 16 different test results (I have 8 scenarios and tested each with a Kolmogorov-smirnov and Shapiro-wilk test), 13 yielded a normal distribution of over .05. So I assumed I should then use an ANOVA test to see if there's significance. However, when I double-checked this with a professor in my department, he (who said he also is no stats expert) said I shouldn't assume normality with any of my results, and stick to using a Kruskal-Wallis test over ANOVA. I'm not really sure why, but I'm wondering if someone could let me know what they think? My sample is 96 people if that has any importance. Thank you!