Wild Cluster Bootstrap command in SAS (regression with few clusters)

[Yuna1]

Hello,

I am trying to run some regressions with clustered data. That is, i have a dataset with about 60,000 observations of unique individuals who belong to about 15 clusters. I would like to correct my regression estimates for the fact that errors are very likely correlated among individuals who belong to the same cluster. Cameron, Gelbach, and Miller (2008) suggest the use of the wild cluster bootstrap procedure to address the situations where there are only a few clusters (working paper: https://www.econstor.eu/bitstream/10419/31404/1/514886757.pdf)

I am wondering if anyone knows the code that should be used to implement the wild cluster bootstrap procedure for regressions in SAS?  I understand that the equivalent Stata code is cgmwildboot (source: https://sites.google.com/site/judsoncaskey/data)

I am using SAS version 9.4 on Windows.

Any thoughts or suggestions are greatly appreciated!

Many thanks in advance,

Yuna

[PGStats]

Please fix your first link

Invalid Identifier

The identifier 10419/31404/1/514886757.pdf).  does not correspond to a valid Bitstream in DSpace. This may be because of one of the following reasons:

• The URL of the current page is incorrect - if you followed a link from outside of DSpace it may be mistyped or corrupt.
• You entered an invalid ID into a form - please try again.

If you're having problems, or you expected the ID to work, feel free to contact the site administrators.

[Yuna1]

Thanks for pointing that out!  I've updated the link (hopefully it works). If not, I've attached the published paper to the original thread.

Thank you!

Yuna

[Rick_SAS]

I haven't implemented the wild bootstrap in any publically available code, but you can see the main ideas of coding bootstrap methods in SAS in a series of 2016-2017 articles at

https://blogs.sas.com/content/tag/bootstrap-and-resampling/

[Yuna1]

Thank you for the additional information!

[sumelioni]

Hi Yuna,

I am currently trying to use the wild cluster bootstrap procedure.  Have you found (or anyone else know of) code to do this in SAS?

Any help would be much appreciated.

[Yuna1]

Hello,

Unfortunately no, I wasn't able to find any specific code for the Wild Cluster Bootstrap procedure.  There may be some information in the links provided by the other users above. So far, I only  know that this capability is built in to Stata, but this may not be very helpful considering we are on the SAS message board.  🙂

I apologize that I'm not able to offer much more help on this!

Best of luck,

Yuna

[sumelioni]

OK, thanks for replying so quickly. I may just have to use Stata to
address this.