Many Stata commands don't leave behind confidence intervals. IMHO, this
is a significant problem (e.g. it took me some time to figure out
exactly how to compute the confidence intervals). However, it is easy to
fix. The program below will display what you want. It will compute
estimates and confidence intervals after svymean and svytotal. In the
other program below this is also written to a file with an option to
have a multiplier (for percentages).
If you read Stata, it should be easy to understand. It gets the
estimate, computes the standard error, and computes confidence intervals
using the t distribution and the number of residual degrees of freedom.
In tests I've run it matches all the digits that I can get from the
confidence intervals printed -svymean- using maximum number of digits.
When I make tables, I loop over each cell in nested for loops (e.g. 11
symtoms by 2 different genders and 4 age groups) and output each cell
nicely with 95% confidence intervals to a text file. No more cut and
paste of hundreds of cells for me! (I described this in an earlier post
which didn't generate any commotion). This program is -not- polished
(uses named scalars, cannibalizes a matrix).
Jan Brogger
program define dispsvymean
version 7.0
foreach scal in est se lo hi {cap scalar drop `scal'}
mat est=e(b)
scalar est=est[1,1]
mat se=e(V)
scalar se=sqrt(se[1,1])
scalar lo= (est-(invttail(e(df_r),0.025)*se))
scalar hi= (est+(invttail(e(df_r),0.025)*se))
di %3.2f est " " %3.2f lo " " %3.2f hi
end
program define writesvymean
*will write out an estimate + conf.int after svymean and svytotal
version 7.0
syntax , MULtipl(real) [ fileh(string) ]
if "`fileh'"=="" {local fileh "myfile" }
foreach scal in est se lo hi {cap scalar drop `scal'}
mat est=e(b)
scalar est=est[1,1]
mat se=e(V)
scalar se=sqrt(se[1,1])
scalar lo= (est-(invttail(e(df_r),0.025)*se))*`multipl'
scalar hi= (est+(invttail(e(df_r),0.025)*se))*`multipl'
scalar est=est*`multipl'
di %3.2f est " " %3.2f lo " " %3.2f hi
file write `fileh' %3.2f (est) _tab /*
*/ %3.2f (lo) _tab /*
*/ %3.2f (hi) _tab
end
-----Original Message-----
Date: Tue, 05 Nov 2002 11:07:02 -0600
From: Suzanne Graden <[email protected]>
Subject: st: svytotal output in fixed format
Good day.
In my analysis using svy commands (svytotal), I need estimates and
confidence intervals with fixed format precision instead of the general
format which results in the estimates and ci displayed in scientific
notation.
I have been able to display the point estimate using "matrix list
e(est),
format()" but I need help to obtain the upper and lower confidence
limits.
When I use "estimates list" the confidence limits do not seem to be
stored.
Is this correct or am I overlooking something? I tried to calculate the
confidence limits by first using the "svmat" command to change the
e(V_ds)
to a variable and then formatted the variable using format(), but the
result was still displayed in scientific notation. What is the best way
to
obtain the confidence limits displayed in fixed format when using svy
commands that do not allow the user to adjust the table cell size?
Suzanne Graden
*
* 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/
