CBT service eval data analysis

[Blondepsych]

Hi,
I am conducting a service evaluation on a manualised CBT for anxiety group, at a college for individuals with autism and associated difficulties for my final MSc project.

The purpose of the evaluation is to investigate the clinical outcome effectiveness of the CBT group.

I have analysed pre and post intervention anxiety scores (from GAS, standardised questionnaire) using a within subjects t-test.

Other variables measured on a 5 point likert scale in each of the 6 sessions for every student included: mood, comprehension, participation and behaviour. I am stuck as to how to analyse this data, any suggestions?

 

[GretaGarbo]

I would just calculate the mean for each Likert scale (and for each Likert item) and do t-tests.

 

[Blondepsych]

Thanks, I used wilcoxon signed rank test as my data did not fulfil parametric assumptions.

For mood, participation, comprehension and behaviour, my results were non significant. Would I report that there was a non significant effect of e.g comprehension on the reduction of anxiety?

I know my questions are very basic, feeling very unsure of myself the closer I am to my deadline

 

[GretaGarbo]

Yes you should report what is significant and what is not significant.

(By the way, there are many parametric test, many more than the t-test.)

Corresponding to a Wilcoxon signed rank test is the paired t-test. I would do a paired t-test on the individual differences (and report the corresponding means). It would probably be easier for your audience if you report means and confidence intervals (although it gives the same result as the t-test, but people find it easier to understand).

 

[Blondepsych]

Thanks so much for your reply.

I was told not to conduct a t test as my data is not normally distributed, and does not fulfill other assumptions required to conduct a parametric test, which is why I conducted the Wilcoxon.


A Wilcoxon signed ranks test revealed that the difference between pre and post intervention GAS scores was non-significant, z= 1.63, p>0.05, with a large effect size (r=0.77). The median score on the GAS decreased from pre intervention (Md= 26.50) to post intervention (Md=24.00).

When it comes to the other 4 variables; mood, participation and behaviour, are you recommending I report the means instead of the medians like so:


A Wilcoxon signed ranks test revealed that the effect of mood on the difference in GAS scores pre and post intervention was non-significant, z=0.73, p>0.05 with a medium effect size (r=0.30). Negative rank (3.00) and positive rank (3.75).

many thanks

 

[Blondepsych]

Or
A Wilcoxon Signed-Ranks test indicated that the effect of mood (mean positive rank = 3.50) on the difference in GAS scores pe and post intervention (mean rank = 3.50), Z = 0.73, p>0.05 was non significant.​