Interpreting/Graphing Continuous Interactions

[Mariame Shaw]

Hi Everyone,

I have a question regarding interpreting one of my interactions! Specifically, after running a linear regression model where the interaction term was significant (B=-.016; p<0.001), I graphed the interaction (e.g., at one standard deviation below and above the mean) using the margins and marginsplot commands. However, the slopes of my graph were parallel (suggesting no significant interaction), albeit my interaction term was statistically significant in the model.
I was wondering if I had missed anything, if anyone had any advice on next steps or resources for interpretation if possible as this is the first time I ran into this issue.

 regress outcome c.c_predictor c.c_moderator c.c_predictor#c.c_moderator c.age ///
            i.sex i.race i.mari i.ra_edu i.emp_status

            
           //Graph at -1sd/mean/+1sd of predictor and moderator
           margins, dydx(c_predictor) at(c_moderator=(-.4828263 0 .4828263))
            margins, at(c_predictor=(-1.869654 0 1.869654) c_moderator=(-.4828263 0 .4828263))
            marginsplot, x(c_predictor)

[Dirk Enzmann]

If I look closely I think that I can see a small interaction -- what I if you use 3 standard deviation below and above the mean? Your b coefficient is relatively small (note that significance does not imply a substantial effect if the sample size is large).

[Mariame Shaw]

Thank you so very much! I have tried using 3 and 5 standard deviations away, and they are both better at representing the interaction. Thank you so very much!

Graph: 3 standard deviations away


Graph: 5 standard deviations away