I have a very basic understanding of statistics, and have done many correlations and regressions or simple t-tests, but never an interaction and it is confusing me a bit. I only need to understand the basics of it for now and would really appreciate any input from all you statistics experts.
I just basically need to understand what an insignificant interaction means, when the main effects and the conditional effects are all significant.
I am using modprobe syntax (Hayes, 2010) for SPSS to test an interaction between 3 variables.
I have 3 variables they are:
Outcome Variable InterVio (Inter Group Violence) (scale from 0-13) Focal Predictor Variable Ruminati (Rumination) (Scale from 1-4) Moderator Variable ColNarc (Narcissism) (Scale from 1-6)
The hypothesis (simple psychology report hypothesis) is: That narcissism and rumination interact at high levels to predict higher aggression and interact to predict lower aggression at lower levels.
The overall model shows significant main effects, but insignificant interaction:
Complete Model Regression Summary R-sq F df1 df2 p n .3726 28.7066 3.0000 145.0000 .0000 149.0000 =================================================================== b se t p constant -.0407 .6950 -.0585 .9534 Ruminati .6450 .2777 2.3229 .0216 <---- ColNarc .5646 .2258 2.5000 .0135 <---- interact -.0690 .0838 -.8237 .4114 <---- Interact is defined as: Ruminati X ColNarc =====================================================================
I quite simply want to report this interaction in a psychology research report (basic one).
I just do not understand how to interpret the insignificant interaction. I think I could say that "the IVs predict the DV independently but the interaction was not significant (therefore it is by chance that at different levels of one IV the second IV and the DV increase?)
But then the output shows 'the conditional effect of focal iv at values of the moderator iv'.
Conditional Effect of Focal Predictor at Values of the Moderator Variable ColNarc b se t p LLCI(b) ULCI(b) 2.2472 .4900 .1169 4.1920 .0000 .2590 .7210 3.2077 .4237 .0903 4.6905 .0000 .2452 .6023 4.1682 .3574 .1249 2.8614 .0048 .1105 .6043 Alpha level used for confidence intervals: .05 Moderator values are the sample mean and plus/minus one SD from mean
My problem is that these are significant and therefore makes me wonder what they mean. Because if the entire interaction is insignificant (occurred by chance, as I understand) how can it still interact significantly at different values? (So would I ignore the interaction significance and just report the conditional effects and the slope plot?)
Also does this table mean that the slope line plot graph would be statisticall significant? (But how can it be if the interaction is insignificant? Its so confusing for me, how anything can be significant if the interaction isnt...)
Basically it would be great if someone could just tell me what the significance of the last table actually is meant to be telling me.
Also I was wondering if marginal effects could mean the same thing as conditional effects?
Edited to Add: I have supposedly enough literature to support this interaction existing, so maybe I should run a mediation instead after I find the insignificant interaction? Sorry for rambling everyone, but Im just getting more and more confused....
Thank you very much for anyone who bothers to even read this. I feel useless at stats sometimes.