I came across a significant difference in computation time between the two lines -margins var1#var2- and -margins var2, at(r(levels)). I use age dummies instead of treating it as a continuous variable (-c.age-), as to plot the crude data.
The full code is the following:
Why is there such a big difference in computation time? Is any of the two 'wrong'? The resulting plots are identical.
The full code is the following:
HTML Code:
use https://www.stata-press.com/data/r18/nhanes2, clear
timer clear
timer on 1
logit highbp sex##age // estimates dummies, at() and margins with interactions is equivalent
margins age#sex // does not work with continuous age (after specifying -c.-) (but -at() also works in this case)
marginsplot, name(g2, replace)
timer off 1
gr combine g1 g2
timer list 1
**can also compute age#sex using at(), but this is much slower, see comparison below**
timer on 2
logit highbp sex##age // estimates dummies, at() and margins with interactions is equivalent
levelsof age
margins sex, at(age=(`r(levels)'))
marginsplot, name(g3, replace)
timer off 2
timer list 2
gr combine g2 g3
Comment