2
$\begingroup$

I have read a few posts on related topics but can't seem to figure out which test would best apply in my study. What I have is ordinal data from the same questions taken at two points(W1 and W2). After the initial rating W1 respondents were exposed to an article (A) and instructed to give ratings to the same question again (W2). I have four groups of people, each was exposed to a different version of the article (A1 to A4). In all four groups the second rating (W2) was lower than the first rating (W1), however, looking at the data I can tell that in some groups this decline was larger. Is there any non-parametric test that would allow me to compare (W1) with (W2) between 4 groups of articles (A1 to A4) ?

So far I used Kruskall-Wallis and there are signifacant differences in W2 between groups A1 to A4, but the problem is it doesn't take into account the fisrt measurenment point.

Friedman test, from what i understand, allows me to compare few measurement points but will no allow me comparison between article groups. Please can anyone help?

Many thanks in advance

Improve this question
$\endgroup$

1 Answer 1

Reset to default
5
$\begingroup$

An easy solution that I have seen done in this situation is to run the same Kruskall-Wallis test that you have been using, but on the difference scores rather than just the W2 scores. So, for each participant, you would have WD = W2 - W1 (or vice versa), with the same grouping variable.

This would answer a question about the differences in magnitude of the effect between groups.

Improve this answer
$\endgroup$
1
  • $\begingroup$ Yes, I have also tried doing this and I agree its probably the easiest solution. Just though there may be something I am overlooking here. Many thanks for response. $\endgroup$
    – George
    Commented Apr 7, 2013 at 0:43

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