Hi,
I am trying to obtain a table displaying proportions of a categorical
variable after multiple imputation using Amelia. I guess what I am looking
for is something equivalent to Stata's svyprop command that gives us
proportions with standard errors. Expressed differently, I would like to
get tables instead of summary statistics (as in misum) within the MI
command family, or be able to request something as simple as 'table' or
'tab' etc., instead of the various more sophisticated regression estimators
offered for 'miest'. I would be grateful if somebody could point me in the
right direction (especially if this turns out to be the wrong list).
Many thanks in advance for your help!
Patrick
--
Clarify mailing list served by Harvard-MIT Data Center
[Un]Subscribe/View Archive: http://lists.hmdc.harvard.edu/?info=clarify
Hi,
I'm trying to run an ordered logit in Stata using Clarify with multiply
imputed datasets.
When I run the ologit just using miest, I have no problem. Once I add the
Clarify wrapper, though, I'm not able to get any results.
I'm using syntax like this:
. estsimp ologit pension income age, mi(imp1)
and I get the error message:
"no variables defined"
Same story if I list the imputed datasets separately instead of giving the
stub.
This sounds like a basic problem, but I'm not an experienced Stata user, so
this doesn't make any sense to me. Help?
thanks.
Julie
Julia Lynch
Robert Wood Johnson Health Policy Scholars Program
Center for Basic Research in the Social Sciences
Harvard University
34 Kirkland St.
Cambridge, MA 02138
tel 617 495-5366
email jlynch(a)latte.harvard.edu
and
Assistant Professor
Dept. of Political Science
University of Pennsylvania
202 Stiteler Hall
Philadelphia, PA 19104
email jflynch(a)sas.upenn.edu
Kelly,
Clarify with the cluster command won't take account of efficiencies gained through stratification-- that's why the standard errors are slightly bigger. (The STATA users manual has a nice discussion of this point.) Here's how I think about it: If your key results do not appear to be sensititive to the use of the stratification variable, use clarify and report that your statistical tests are "conservative" because standard error estimates are larger than they would otherwise be.
I'd be interested to know whether others have a different take on the issue.
-- Katherine
Katherine M. Harris, Ph.D.
Office of Applied Studies
Substance Abuse and Mental Health Services Administration
5600 Fishers Lane, Rm 16-105
Rockville, MD 20857
Phone: 301.443.0747
Fax: 301.443.9847
email: kharris(a)samhsa.gov
>>> Kelly Richardson <Kelly-Richardson(a)uiowa.edu> 02/09/04 01:19PM >>>
Hello,
I'm analyzing the AddHealth data set and was planning to use multiple
imputation to handle the missing data. According to the AddHealth
methodologists they recommend the following Stata syntax to correct for the
complex survey design.
svyset [pweight=gswgt2] , strata(region) psu(psuscid)
svynbreg depvar indvars, subpop(sat_schl)
I've seen a couple messages in the Clarify archives regarding clustering
and weights but I wasn't sure if the same approach would work to account
for everything in the syntax above. In addition to weights, stratification
and clustering, the syntax also uses the "subpop" command to adjust the
standard errors when using a subpopulation. I've run my own analyses
(without multiple imputations) comparing the results using the syntax above
to what I would get if I just did nbreg and used the weights and cluster
commands. The coefficients are pretty much the same and the standard errors
are slightly smaller using svynbreg. Is it ok to just use estsimp nbreg
since Clarify (or MI) can't be used with svynbreg? Are there additional
adjustments I should make if I do so? If you don't recommend using Clarify,
do you know if there are any programs that can be used with the survey
regressions in Stata?
Thanks in advance for any help you can provide.
Kelly
Kelly Richardson
Department of Sociology
W140 Seashore Hall
University of Iowa
Iowa City, IA 52242
--
Clarify mailing list served by Harvard-MIT Data Center
[Un]Subscribe/View Archive: http://lists.hmdc.harvard.edu/?info=clarify
--
Clarify mailing list served by Harvard-MIT Data Center
[Un]Subscribe/View Archive: http://lists.hmdc.harvard.edu/?info=clarify
Hello,
I'm analyzing the AddHealth data set and was planning to use multiple
imputation to handle the missing data. According to the AddHealth
methodologists they recommend the following Stata syntax to correct for the
complex survey design.
svyset [pweight=gswgt2] , strata(region) psu(psuscid)
svynbreg depvar indvars, subpop(sat_schl)
I've seen a couple messages in the Clarify archives regarding clustering
and weights but I wasn't sure if the same approach would work to account
for everything in the syntax above. In addition to weights, stratification
and clustering, the syntax also uses the "subpop" command to adjust the
standard errors when using a subpopulation. I've run my own analyses
(without multiple imputations) comparing the results using the syntax above
to what I would get if I just did nbreg and used the weights and cluster
commands. The coefficients are pretty much the same and the standard errors
are slightly smaller using svynbreg. Is it ok to just use estsimp nbreg
since Clarify (or MI) can't be used with svynbreg? Are there additional
adjustments I should make if I do so? If you don't recommend using Clarify,
do you know if there are any programs that can be used with the survey
regressions in Stata?
Thanks in advance for any help you can provide.
Kelly
Kelly Richardson
Department of Sociology
W140 Seashore Hall
University of Iowa
Iowa City, IA 52242
--
Clarify mailing list served by Harvard-MIT Data Center
[Un]Subscribe/View Archive: http://lists.hmdc.harvard.edu/?info=clarify
