Kit has just posted an important update to boottest. Important, because it fixes a bug that affected reported results when using observation weights (pweights, etc.) I'm very sorry for this error.
In addition, the new version:
In addition, the new version:
- Makes the Kline & Santos score bootstrap available after LIML estimation with ivreg2 and ivregress.
- Computes confidence intervals by default. This applies when estimation is analytical (OLS, 2SLS, GMM, LIML) rather iterative (Maximum Likelihood) and the null hypothesis is a single constraint. The option "ci" has been dropped and replaced with a "noci" option that increases speed when you don't need the CI.
- Takes a new option, svmat, so you can capture and view the full empirical distribution, as recommended in Cameron & Gelbach's "Practitioner's Guide." The help file has an example of graphing the distribution.