2
$\begingroup$

For example, when I compared a student population split into four year groups and compiled a bar graph, I could see that the 2nd year students were vastly overrepresented in their responses to that particular question.

Seperately, I performed a Kruskal Wallis test to see if there were any statistically significant differences between the year levels, which returned a p-value of <0.05. A quick scan of the internet has told me that I need to do a post-hoc test to figure out where the significant difference lies. But it would seem that I already know the answer to that question? Could I not just say "The larger number of 2nd year responses over all other year groups was statistically significant"?

Cheers.

$\endgroup$
2
$\begingroup$

Significance testing has sometimes been criticized for only telling researchers things they already know. I think there's some merit to this criticism. In any case, though, it doesn't make sense to do significance testing halfway. Either carry it through to its logical conclusion, which in this case seems to call for a post-hoc test, or don't use it at all. After all, if your eyeballs are sufficient evidence that the second-year students responded more than the other students, what's the point of the initial Kruskal–Wallis test?

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.