Plotting results of regression with interaction terms

kjo

New Member
#1
This is tangentially R-related, but mostly about presenting results of a regression including interaction terms. My ultimate objective is to produce a coefficient plot representing differential associations between a set of predictors and responses for two groups. To this point, as an alternative to bivariate regression or just running separate regressions for each group, I've run a regression that interacts all IVs with a dummy-coded "groupid" variable. I've plotted the results using coefplot2 in R, but would like plots grouped by each IV to better represent the differential association for each x and y. I can do this by hand using ggplot2 if I have to, but I'm not 100% sure about which points to plot, especially for the min and max of the confidence bands. With the example, Y = b0 + b1(X) + b2(Z) + b3(XZ) + e, plots for Group 1 (Z=0) will be centered at b1 with min and max for the confidence band formulated as b1+-1.96*SE. Plots for Group 2 (Z=1) will be centered at b1+b3, correct? If that's right, how do I formulate the min and max for the confidence band for Group 2? Would I just use the SE associated with b3? Again, the end result is meant to be a coefficient plot representing the association between each predictor and response for both Groups 1 and 2.