Can I keep continuous variable as the moderator in a moderation model? The regressor variable is categorical variable. I am also interested in 2-way and 3-way interaction modelling. In both the cases, moderating variable are continuous whereas regressor variable is the categorical variable.
The dependent variable is continuous as well.
Sure, why not? You mention Stata. Study up on -margins- and -marginsplot- to see how to make plots that show the nature of your interaction. Here's a silly example using the auto.dta data file that comes with Stata.
clear *
sysuse auto
tab rep78
drop if rep78 == 1
summarize weight displacement
regress mpg rep78##c.weight##c.displacement
margins rep78, at(weight=(2000(1000)4000) displacement=(100(100)400)) vsquish
marginsplot
graph rename g1
marginsplot, bydim(displacement)
graph rename g2
marginsplot, bydim(weight)
graph rename g3
HTH.
Assuming that you have a factorial experiment that is large enough to deal with your 3-way interactions I see no problem as long as the number of replications is also large enough. For details see:
https://www.google.com/search?q=Montgomery+Design+of+experiments&rlz=1C1CHBF_enUS874US874&oq=Montgomery+Design+of+experiments&aqs=chrome..69i57.23560j0j1&sourceid=chrome&ie=UTF-8
Best , D. Booth
Sure, why not? You mention Stata. Study up on -margins- and -marginsplot- to see how to make plots that show the nature of your interaction. Here's a silly example using the auto.dta data file that comes with Stata.
clear *
sysuse auto
tab rep78
drop if rep78 == 1
summarize weight displacement
regress mpg rep78##c.weight##c.displacement
margins rep78, at(weight=(2000(1000)4000) displacement=(100(100)400)) vsquish
marginsplot
graph rename g1
marginsplot, bydim(displacement)
graph rename g2
marginsplot, bydim(weight)
graph rename g3
HTH.
- Badges
- Science topic