I'm trying to investigate the similarity of driving behaviour between two types of vehicle technology. I took a Mannwhitneyu test and the p-value shows that the behaviour is different, whereas when I fit a distribution to the dataset, they are pretty much similar (alpha and beta for Weibull function). The data set is huge also. I'm not sure I'm doing the right thing for comparing them. Any references or suggestions I could refer to?
It sounds like it's an effect of having a large sample size. Statistical tests can report small p-values for large samples even when the effect size is small. My advice: Don't rely too heavily on the p-value. Instead, assess the effect size (for example with an effect size statistic or, say, the difference in medians), and the practical importance of the effect. E.g., would this difference matter to anyone?
It is an interesting question, I have gone through above answers, my opinion is increase or decrease the sample size randomly(I.e change the sample) and perform the exercise once again. Report the result, I think you may get the answer.
- Badges
- Science topic
- Similar topics
- Psychology
- Mental Processes
- Learning