In response to a query from Jérôme Valette and Sarah Schneider-Strawczynski, I've added an interesting feature to boottest, the ability to bootstrap results from margins. I had never thought about doing this, but some reflection led to the realization that it could be easy because most (all?) results from margins are linear combinations of the primary parameters, and the coefficients of those combinations are returned by margins in r(Jacobian).

To use the new ability, simply run a margins command, then do "boottest, margins" and it will produce bootstrap p values and confidence intervals separately for each marginal effect. Please let me know if you run into problems.

Examples:
Code:
use nlsw88
regress wage tenure ttl_exp collgrad south#union, cluster(industry)
margins south
boottest, margins
margins, dydx(south)
boottest, margins graphopt(xtitle(Average effect of south))


The new version is available via
Code:
net install boottest, replace from(https://raw.github.com/droodman/boottest/v3.2.0)
I will have it posted on SSC soon.