Any one have experience in mediation (SPSS)

Anything that does not fit into the above categories, but is related to psychology, including discussion of public and media perceptions of psychology, satire related to psychology, etc.
Post Reply
DrFurbs
Posts: 82
Joined: Fri Mar 27, 2009 5:51 pm

Any one have experience in mediation (SPSS)

Post by DrFurbs » Sat Feb 27, 2010 4:32 pm

IV > Mediator > DV?

I have an issue, and Im wondering If I could ask a few questions.

Ruthie
Moderator
Posts: 3100
Joined: Sat Mar 24, 2007 11:32 pm
Location: London
Contact:

Post by Ruthie » Sat Feb 27, 2010 10:29 pm

Fire away! I'd rather you asked your question on the public forum so that other people can benefit from the answers afterwards. I have experience with mediation in SPSS, but I am sure I'm not the only one.

DrFurbs
Posts: 82
Joined: Fri Mar 27, 2009 5:51 pm

Post by DrFurbs » Sun Feb 28, 2010 2:21 am

Thanks Ruthie.

I'm talking specifically about the Preacher and Hayes SPSS bootstrapping macro script which does away with the Baron and Kenny causal steps approach in H.Regression. The output of the script is complicated as it does not rely on p values etc for determining significant direct and indirect effects, but confidence intervals and point estimates.

I should have made that clear on my first post (stupid me).

http://www.comm.ohio-state.edu/ahayes/S ... direct.htm

Basically my problem is that the CI are not significant in the output, thus the mediators in my model anxiety and depression do not mediate the relationship between the IV and DV which is contrary to what I expected. Im wondering how on earth am I going to explain this in my dissertation as 1 of the prerequisites for this script and model based on theoretical grounds is that the variables are correlated to some degree, which they do in the literature.

Childhood bullying experiences (IV) > (M) anxiety + (M1) Depression > Muscle Dsymorphia (DV)

Penny for your thoughts. Btw heres the output if you want to read it.

http://img132.imageshack.us/img132/7690 ... hackeh.png

The bootstrap results contradict the Normal theory Tests for indirect effects. Basically the Bootstrap area says nothing is significant as zero lies between the confidence intervals and the normal theory tests says anxiety is significant...but its the BS area that I must adhere to.

so, how do I explain this :)

Ruthie
Moderator
Posts: 3100
Joined: Sat Mar 24, 2007 11:32 pm
Location: London
Contact:

Post by Ruthie » Sun Feb 28, 2010 11:58 am

:shock: It sounds like you have a very good grasp of mediation!

It looks like you have just missed significance for a mediating effect of anxiety in your BS test. Double check for any outliers or quirks in your data set. You could also re-run the analysis with just anxiety as a possible mediator as it looks like depression is definitely not a mediating factor. I'm not entirely sure how BS works, but I'm wondering whether the association between anxiety and depression is effecting your results? Maybe check with your supervisor or someone more knowledgeable.

Either way, the results suggest the mediating effect is pretty small and that is probably the important thing to note. There is some mediation, but it isn't a big effect size. How you interpret that is down to the theories you're using!

DrFurbs
Posts: 82
Joined: Fri Mar 27, 2009 5:51 pm

Post by DrFurbs » Sun Feb 28, 2010 9:54 pm

Hi Ruthie, I would have a better grasp if this was H.Regression. :(

Any ways, I had a look for outliers and indeed found some including 1 extreme case. Was was I to do in this case? I thought I could delete said cases but opted for transforming the variables in question to improve the distribution. However, although this has vastly improved the dataset, im reading there is some controversy surround this technique and to be brutally honest, I'm not even sure if this should be done considering the data was boot strapped 5000 times (Bootstrapping is a technique from which the sampling distribution of a statistic is estimated by taking repeated samples (with replacement) from the data set) so im unsure if im cheating or breaking rules so to speak.

However, the CI of anxiety NOW agree with the normal theory of tests, although the CI for depression still does not BUT!!! This would coincided with the theory that those who were depressed feel lack of drive to change themselves and lifestyle, lack of energy and above all motivation, hence depression would not be a mediator for a body distortion disorder, where as anxiety has been shown in the literature to be significant in this respect.

Time to see my professor!

DrFurbs
Posts: 82
Joined: Fri Mar 27, 2009 5:51 pm

Post by DrFurbs » Mon Mar 01, 2010 4:37 pm

Ruthie, good news. What I did was completely valid. Thanks for the heads up in terms of outliers.

Time to finish this dissertation.

Post Reply

Who is online

Users browsing this forum: No registered users and 1 guest