Insignificant interaction effect in regression vs. scatterplot indicating interaction

#1
Hallo,

I'm new to this forum and hope I can make my problem clear.

I conducted a multiple regression analysis to, among other things, test the moderating effect of environmental concern on the relationship between the sustainability of the packaging material and perceived quality.

Sustainablity of the packaging material is dummy coded (1=paper packaging, 0=plastic packaging). Environmental concern is a continous variable.

The regression analysis indicates that there is no moderating effect of environmental concern as p=.314.

However, if I regress environmental concern on perceived quality it appears that there is an interaction effect (see attachment).

I'm asking myself how this is possible as the visual effect seems quite significant. The data for environmental concern are very unbalanced. Most of my participants were highly environmentally concerned, meaning I only have a few people stating they are low environmentally concerned. Is it possible that the unbalanced sample for this variable explains the insignificance of the moderating effect in the regression?

Thank you in advance,
Nora
 

hlsmith

Omega Contributor
#2
Re: Insignificant interaction effect in regression vs. scatterplot indicating interac

Are the line crossing outside of the exiting data (extrapolation when interpreting the lines)? So if I draw a line from the top left blue dot down to the x-axis, the intersection of the two slopes occurs on the left-side of the lowest value data point!