cgmreg and cgmwildboot

[Christoph Engel]

This is just to help others who might face the same problem. Cameron Gelbach & Miller RevEcStat 2008, 414 explain why standard errors tend to be too narrow when one clusters, and the number of clusters is small. One way out is the standard bootstrap. But CGM explain why the wild bootstrap is even more reliable. In principle, user written software is available. But one cannot install it with the usual findit routine. The following is what has worked for me

1. cgmreg (from Jonah Gelbach himself)
The ado-file is available on his website (albeit somewhat hidden)

http://gelbach.law.upenn.edu/~gelbach/ado/cgmreg.ado

By cut and paste one can put this into Stata's do-file editor, and then save (as an .ado file!) at C:\ado\plus\c

Then Stata finds it. It seems there is no help file (but at the beginning of the .ado file, the logic of the command is explained)


2. cgmwildboot (from Douglas Miller and co-authors)

By using

findit cgmwildboot

one in principle can install it. Yet the command requires the

unique

command, which is currently also not available through findit. It can, however, be found at

http://fmwww.bc.edu/repec/bocode/u/

One must also separately download the .hlp file, and save both at C:\ado\plus\c

With best regards

Christoph Engel

[Michael Fryar]

Thanks, Christoph! One small note, unique.ado and unique.hlp must be saved at C:\ado\plus\u.

Best,
Michael

[Chen Samulsion]

The URLs below are both invalid:
http://glue.umd.edu/~gelbach/ado/cgmreg.ado
http://gelbach.law.upenn.edu/~gelbach/ado/cgmreg.ado
people who are interested in Two-way Cluster Robust Standard Errors should try this link in A. Colin Cameron's website:
http://cameron.econ.ucdavis.edu/research/papers.html

[Gabriel Facchini]

Is there anyway to export the CI from cgmwildboot? can't find them in the save results. thanks!