[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: BOOTSTRAP and pweights
This is a repost, I originally forgot the subject like.
Sabrina Di Addario <SABRINA.DIADDARIO@bancaditalia.it> asks about using
-pweight-s with -vce(bootstrap)-:
> I am trying to use bootstrap now (rather than jackknife) to do the same
> as before:
> biprobit (y1= x2 y1 y2 u1) (y2 = x2 z2) [pweight=weight], robust
> I tried several specifications - none of which worked:
> 1) biprobit (y1=x1 x2 y2 u2) (y2 = x2 z2), ///
> vce(bootstrap [pweight=weight]) cluster(clusty)
> 2) biprobit (y1=x1 x2 y2 u2) (y2 = x2 z2) [pweight=weight], ///
> vce(bootstrap) cluster(clusty)
> 3) bootstrap, reps(100) seed(1): biprobit (y1=x1 x2 y2 u2) (y2 = x2 z2)
> [pweight=weight],/// robust cluster(slnew)
> Is there a similar problem as in jakknife in bootstrap too or there is
> another way of writing the syntax? Or, more simply, bootstrap does not
> allow pweights?
Although you may get wierd error messages with syntax 2 and 3 (which will be
fixed in an upcoming ado-file update), -bootstrap- does not allow weights as
reported by Stata if you use syntax 1.
Question: Why doesn't Stata allow weights with -bootstrap-?
Besides the book by Shao and Tu (1995), there are papers in the survey
literature on using the Bootstrap with complex survey data. Unfortunately
there doesn't appear to be a single satisfactory method for Bootstrapping
data with sampling weights.
Given the number of alternative methods for bootstrapping weighted data and
the lack of a clear winner (popular or otherwise), we decided to disallow
even the 'naive bootstrap' until there was sufficient demand (at least for a
statistically satisfactory method).
The only exception here is when you have bootstrap replicate weights. You can
-svyset- the bootstrap replicate weights as -jkrweights()- with the proper
-multiplier()- option (or use -brrweights()- with the proper -fay()- option).
Shao, J. and D. Tu. 1995. The Jackknife and Bootstrap. New York: Springer.
* For searches and help try: