I seem to be having a problem using Bill Sribney's 'mlogpred' command
after 'svymlog'. I am analysing the 2001 National Drug Strategy
Household Survey (Australia) using multinomial logistic regression and
would like to get 95% CIs for the predicted percentages of persons
drinking at short-term risky and high risk levels by industry of
employment adjusting for age, gender etc. The variable 'shorttmYMW' is a
person's alcohol consumption at levels that are harmful in the
short-term and has the categories 0 "abstainer" 1 "low risk" 2 "yearly
risky and high risk" 3 "monthly risky and high risk" and 4 "weekly risky
and high risk".
When I use a mlogit model that does not have any svy prefix, the
'mlogpred' command works. I get a plausible results for the percentages
of persons drinking at weekly risky and high risk levels by industry of
employment, but it does not take into account the survey's
stratification and clustering.
xi: mlogit shorttmYMW i.industry age age2 sex if work==1
mlogpred weekly, outcome(4)
mlogpred ci1 ci2, ci outcome(4)
tabstat weekly ci1 ci2, by(industry)
Summary statistics: mean
by categories of: industry
industry | weekly ci1 ci2
---------------+------------------------------
agriculture | .1077441 .0933308 .1240925
mining | .0661765 .0522896 .083572
manufacturing | .10139 .0933151 .1100906
construction | .1055684 .0960305 .1159512
wholesale | .0859729 .0720848 .102313
retail | .0933148 .085793 .1014461
transport | .0966719 .0872089 .1070862
financial | .0794507 .0743566 .0848808
administration | .0630303 .0577801 .0687462
services | .0604982 .0570801 .0641256
hospitality | .1533613 .1335594 .1754044
education | .0402145 .0376993 .0429048
---------------+------------------------------
Total | .0809734 .0742147 .0883615
----------------------------------------------
When I repeat the same commands, but first svyset my data and then use
the 'svy:' prefix for mlogit, I get the error message 'last estimates
not found'. The help file for 'mlogpred' states that 'mlogpred' can be
used after svymlog. I am using Stata/SE 9.2 and it is up-to-date. To
check the error was not due to the new 'svy' syntax in Stata 9, I tried
the old 'svymlog' syntax in Stata 8 but got the same result. Thanks in
advance for any help. Jesia
svyset [pw=wts9], strata(areas) psu(newpsu)
xi: svy, subpop(work): mlogit shorttmYMW i.industry age age2 sex
mlogpred weekly2, outcome(4)
last estimates not found
r(301);
*
* 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/
