I have a dataset with 2 cross-sectional independent samples - the
first survey conducted in 1997 & the second survey conducted in 2007.
The var indicating survey year is nsmhwb - coded (0/1).
My svyset statement, based on the 90 jackknife replicate weights etc,
is as follows:
-svyset [pweight=mhsfinwt], jkrweight(wpm0101 - wpm0190,
multiplier(1)) vce(jackknife) mse-
I want to compare proportions on a DV (mhpsyo12 = 0/1) by IV survey
year (nsmhwb = 0/1).
-svy jackknife, nodots : proportion mhpsyo12, over(nsmhwb)-
The 1997 estimate is 17.0% (SE = 1.3%)' the 2007 estimate is 37.3% (SE = 2.5%).
I have separately created a variable that indicate the proportion of
2007 in each of 5 age group strata (i.e., 20-29 years, 30-39 years,
etc to 60-69 years). The strata var is 'st_age)'; the 2007 strata
population proportion var is 'st_wt' - ranging from 0.0 to 1.0.
The population structure has changed over the 10-year period (i.e.,
aged). So, I want to compare proportions adjusted by the 2007 strata
population size I run:
svy jackknife, nodots : proportion mhpsyo12, stdize(st_age)
stdweight(st_wt) over(nsmhwb)
The 1997 estimate is 15.5% (SE = 1.4%)' the 2007 estimate is 34.1% (SE = 2.3%).
I'm puzzled that the 2007 proportion estimates change from the
previous statement. I would have thought that the 2007 estimates (and
their SEs) would NOT vary given the fact that standardization is based
on actual 2007 proportions?
I guess I'm doing something wrong - any ideas?
I'm using Stata 11.0 - born 21 October 2009.
Thanks;
Philip
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
