Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

RE: st: RE: svy probit/logit iv


From   "Zhu, Liye" <lzhu@mail.smu.edu>
To   <statalist@hsphsun2.harvard.edu>
Subject   RE: st: RE: svy probit/logit iv
Date   Wed, 12 Jul 2006 02:17:06 -0500

Thank for Scott and Professor Williams's answers and suggestions! They are very helpful. 

 

I just updated my Stata  to 9.2. Now I can use -ivprobit-. But as Professor Williams said, -ivprobit- cannot be used after svy. I tried both suggestions by Professor Williams. (1)Changing -probit- and -regress- in the -ivprobit- ado file into -svyprobit- and -svyregress-, and saving it as -svyivprobit-. But the program cannot run correctly. The error message is "data not set up for svy, use svyset could not find initial values r(498)". Actually I already did - svyset [pweight=wtper2], strata(stratawr) psu(psuwrx)-. I set the trace on, but still could not figure out what is wrong with the code. (2) adding [iw=wgtvar] in -ivprobit-, and do -estimate store- and -suest-. However, -suest- cannot work with svy either. (-suest m1, svy- doesn't work, but -suest ml- can work). I am guessing that -suest- after -ivprobit- is to provide robust estimation, am I right? If so, why not just choose robust option in the original -ivprobit- command? Can you explain more about the purpose of -suest m1, svy-? Thank you. 

 

I am using a survey data which recommend to use -svyset [pweight=wtper2], strata(stratawr) psu(psuwrx)- in Stata in order to get relative conservative standard error. So, If I cannot use -ivprobit- with -svy-, what's your suggestion to deal with the survey data in this case?

Another issue is, when using -ivprobit-, how to do the endogeneity test and iv validity test (orthogonality and relevance tests)? For the endogeneity test, I tried -hausman- after -ivprobit-, but got the results saying "V_b-V_B is not positive definite". So how can I improve it? Is there any Durbin or Wu flavor of hausman test (similar in ivreg2)? The last line in the result of -ivprobit- is "Wald test of exogeneity (/athrho = 0): chi2(1) = 68236.00 Prob > chi2 = 0.0000". How to interpret this results? For the iv validity test, do you have any suggestions? I know -ivreg2- in Stata 8 can do the Hensen's J and C test, how about -ivprobit-?

Thanks a lot!

Liye


________________________________

From: owner-statalist@hsphsun2.harvard.edu on behalf of Richard Williams
Sent: Tue 7/11/2006 8:57 PM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: RE: svy probit/logit iv



At 08:09 PM 7/11/2006, Scott Merryman wrote:
>-help ivprobit- ?
>
>-ivprobit- can handle pweights and cluster().  I do not work with survey
>data, but doesn't this cover 2 of the 3 main characteristics of survey data
>([svy] page 6)?
>
>Scott

Sounds like a pretty reasonable guess!  However, ivprobit was added
in Stata 9.  Life would be much simpler for Liye if s/he had Stata 9;
besides adding ivprobit it also added much better svy support.  For
many commands, all you have to do is tack on an svy prefix, e.g. svy:
regress .... And, even when the svy: prefix does not work (as appears
to be the case with the ivprobit command) you can usually still get
what you want with iweights and suest, e.g. do something like

ivprobit .... [iw=wgtvar]
estimates store m1
suest m1, svy

Assuming Liye is stuck with 8.2, I suggest typing -findit ivprob- and
install those routines; according to the help files, they are
improved versions of probitiv and tobitiv, and (unlike the originals)
give correct standard errors.

I'm guessing Liye could then make a copy of the program and change
the various probit and regress commands to svyprobit and
svyregress.  Alternatively, I bet Liye could hack the program to
allow iweights, and then use the suest command, e.g.

myivprob .... [iw=wgtvar]
estimates store m1
suest m1, svy

I caution that my guesses might be wrong; there is a lot of matrix
algebra in ivprob and I don't know for sure that it will still work
ok with these hacks.

If you are doing much svy analysis, get Stata 9 if at all
possible.  Or, at least try to double-check some of the results from
your hacked programs with Stata 9.


-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
FAX:    (574)288-4373
HOME:   (574)289-5227
EMAIL:  Richard.A.Williams.5@ND.Edu
WWW (personal):    http://www.nd.edu/~rwilliam
WWW (department):    http://www.nd.edu/~soc

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


<<winmail.dat>>




© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index