Confusingly simple? Or just confused? PLEASE HELP!

#1
I am examining perceived social support and coping in a group of approximately 40 participants (about 20 participants in two groups) who all recieved an intervention to address loss, with data collected at pre (T1), post (T2), one month post (T3), and three months post (T4). I also have demographic information and loss history data.

Regarding perceived social support, I am interested in differences across the four data collection points for 4 subscales. Similarly, I am interested in differences across the four data collection points for 7 subscales.

My hypotheses are simply:
H0: subscale score mean from T1-T4 will remain constant.

HA: subscale score means from T1-T4 will will not remain constant.

I originally planned on using non-parametric techniques but was advised to run a repeated measures analysis.

Despite having very little guidance, I ran the repeated measures analysis for the whole group across the 4 time points for both measures, using all 40 participants. I only found significance on one subscale for coping.

:confused:Questions:
Was I correct in running the analyses for all 40 participant given that I was not looking at differences between two groups?

Is there another statistical analysis approach that would have been more advisable?

Is there a way to examine the signifcant finding for the one coping subscale?

I may just be feeling lost because I did not hypothesize anything other than the intervention would have an effect upon participants coping and perceived social support. .:shakehead

All suggestions and help is much appreciated
 

Karabiner

TS Contributor
#2
I originally planned on using non-parametric techniques but was advised to run a repeated measures analysis.
If you are just interested in the development of scores over time,
without looking for group effects, the "nonparametric" Friedman
test could be used.
Is there a way to examine the signifcant finding for the one coping subscale?
First of all, did you perform 11 analyses? In that case it is usually
advised to correct the significance level for multiple testing.
I may just be feeling lost because I did not hypothesize anything other than the intervention would have an effect upon participants coping and perceived social support
Without control group, you can hardly claim that the intervention had an effect.

Kind regards

K.
 
#3
Thanks for the reply. I did perform 11 analyses. I used Bonferroni, but after consulting a bit decided to use Sidak.

Thoughts about use of Sidak over Bonferroni?

Also, given the nature of the data (repeated measures, no control group, intervention was not medication/medical trial) I believe I have justification for using LOCF to deal with my missing data. Any thoughts about how to build a solid case in this area?