Problems with the interpretation of a moderation analysis

#1
Hi there,

I am undertaking a moderated regression analysis for an experiment I have conducted using SPSS 20.(With an add-on developed by Andrew Hayes for running moderation tests).

I wanted to find out whether the duration of meditation practice leads to reductions in levels of stress between 2 time points. Therefore I used 3 variables to undertake the moderation test.

X = baseline level of stress (pre meditation)
Y = post intervention level of stress (post meditation)
M = duration of meditation practice

The main interaction shows a non significant effect - which suggested to me that any change in the relationship between x and y isn't moderated by M.

BUT then when I look at the simple slopes analysis - the p values for each level are significant at <.05.- suggesting the relationship between x and y changes at low mean and high values of the moderator.

The Johnson Neyman technique also seems to support the latter
interpretation of significance at low mean and high values of the
moderator.

My questions are therefore A) are the contradictory outcomes from the interaction and simple slopes analysis possible if the test has been run correctly?

B) How should this type of outcome be interpreted if the test has been run correctly?

C) If the contradictory outcomes are likely to have been caused by an error, how can I correct this problem?


Please excuse my limited knowledge on this subject as I have not undertaken any test of this kind before.

Many thanks.
 
#2
I have the exact same issue in my moderation analysis. Any information you can provide on if and how you reached a resolution would be much appreciated.