Thanks to Kit Baum, an update to the -robreg- package is available from SSC. To install the update, type

Code:
. ssc install robreg, replace
or use the adoupdate command.

robreg provides a number of robust regression estimators such as MM regression.

Main changes:
  • The new version is a complete rewrite; the previous version is now available as robreg10.
  • robreg supports two additional estimators, robreg ls for least-suqares fits (equivalent to regress) and robreg q for quantile regression (similar to qreg)
  • robreg now supports predict with a variety of options (e.g. predict robust residuals or influence functions)
  • there is a now command called robreg hausman to compare results between different estimators (Hausman tests)
  • performance of robreg s, robreg lts, robreg lqs, and robreg lms has been much improved by using fast nonsingular subsampling algorithms
  • robreg mm can now be refitted with different efficiencies without having to re-estimate the time consuming S estimate
  • robreg lts now uses local improvement and final refinement and thus provides results that are much more stable than in the previous version
  • weights are now allowed by all estimators; all estimators can now be fitted without predictors or without constant
  • the estimators now use a common framework for the estimation of (robust) standard errors based on influence functions; sampling weights and clustering are supported; robreg ls, robreg q and robreg m additionally support the svy prefix (svy is currently not supported by robreg s and robreg mm due to a specific limitation of svy; support will probably be added once a Stata version becomes available in which this limitation has been removed) (estimation of standard errors is not supported for robreg lts, robreg lqs, and robreg lms)
ben