The BDI-II is given to 30 participants prior to treatment. The data is kept in both raw form (interval) and organized in ranks from "minimal, mild, moderate, and severe" (1, 2, 3, and 4, respectively). a 16 week treatment plan is conducted and the BDI-II is repeated. No dropouts. No control group. What method of analysis would be best for this pre-test/post-test design?
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
1
I'd suggest choosing the measurement level for this questionnaire based on literature in your field. As far as I know, the BDI-II is often considered as interval scaled.
Therefore, check if your raw data meets the assumptions of a dependent samples t-test. This test is for a within-subjects design and requires the differences (post-pre) to be normally distributed. You can investigate this with a Q-Q plots and the Shapiro-Wilk test.
If the differences are not normally distributed or you want to test your rank data then use the non-parametric Wilcoxon signed-rank test. However, the dependent samples t-test has higher power than the non-parametric alternative.
-
$\begingroup$ The signed-rank test is virtually as powerful as the $t$-test even under normality and will be more powerful in most situations. I don't entertain the $t$-test in this situation. Also I'd like to know the original thinking behind the creation of the 1-4 grouping. This is incredibly ill-conceived. $\endgroup$ Commented Oct 2, 2015 at 12:29