I need to compare the answers to a questionnaire (pre / post). The questionnaire was designed using a Likert scale. The two groups were made up of populations, some of whom were in both groups but also some were in one group but not the other. The numbers in both groups were small (the research was qualitative). Bearing in mind the difference in population in both groups, which would be the right test to run to compare the results from group a (pre) to group b (post). A Wicoxon rank sum test? Any pointers very gratefully received.
$\begingroup$
$\endgroup$
3
-
$\begingroup$ Please clarify your 3rd sentence and your "Bearing in mind" statement. Do you want simply to compare 2 groups or to work in some additional consideration of underlying characteristics? $\endgroup$– rolando2Commented Feb 27 at 12:52
-
$\begingroup$ Hi, I want to be able to compare the two groups, but because they are not made up of either a/ completely different participants, or b/ have exactly the same participants, I am struggling to identify the appropriate test $\endgroup$– HeatherCommented Feb 27 at 12:56
-
$\begingroup$ maybe mixed-effects model with a random intercept by subject ID $\endgroup$– Patrick CoulombeCommented Feb 27 at 13:58
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
Unfortunately there is no test that will strictly apply. Your data do not satisfy the assumptions/requirements of either independent-groups tests or dependent-groups (paired-scores) tests. There is no null hypothesis that concerns the way chance will produce scores when people sometimes sit in one group and sometimes in both. You can forego inferential statistics and simply provide descriptive results. Or you can select from your data so that you have either a subset of persons who fit the independent-groups arrangement or a subset that fits the dependent-groups arrangement.