Matt Keough posted on Saturday, April 11, 2015 - 12:27 pm
I am looking at how a two way interaction predicts maturing out of alcohol use as students leave university. The overall linear slope for alcohol use is negative, as expected. When I condition my model on the levels of my moderator, I see that my predictor of interest has a negative regression weight for predicting change in alcohol use at low levels of the moderator. I take this to mean that the predictor (at low moderator) is associated with a less steep negative slope (less maturing out). Is this the correct interpretation? Can we interpret the regression weight for the predictor on the slope or do we also need to consider the intercept of the slope?
Second, does one need a slope to have significant variance to predict change?
A negative influence on the slope means that the slope gets lower when the predictor increases. That is not affected by the mean slope begin negative - it just gets more negative (so a steeper negative development). You don't need to consider the slope intercept.
No. If you have significant predictors of the slope that already says that the slope varies significantly.