I'm estimating a mediation model using gsem: I post a toy version below. I want to use boottest to give me clustered standard errors (that will involve the bootcluster(x) option, but we don't need that to illustrate the problem). Following estimation, boottest allows me to test hypotheses about individual coefficients, or additive combinations of coefficients (the first three boottest commands, below), but it won't let me multiply them, which is what I need to do to get a mediated (indirect) effect. Will boottest not do non-linear combinations, or am I missing something?

sysuse panel, clear

gsem (gdp_cap_l <- train rr i.year i.id) (train <- rr ep i.id i.year)

boottest _b[gdp_cap_l:train]=0
boottest _b[train:rr]=0
boottest _b[train:rr] + _b[gdp_cap_l:train]=0

boottest _b[train:rr]*_b[gdp_cap_l:train]=0