st: Error in esttab

[Marlous Tuithof <MTuithof@trimbos.nl>]

Dear statalist, 

I have been using esttab for a while as it is a great way to quickly generate tables. However, I recently noticed that the confidence intervals (and p-values) in my esttab table are slightly different from those in the regression output. I used the following command (stata version 12.1, windows) with multiply imputed data: 

eststo: mi estimate, post or: svy: logistic yypaud xedu_dum yr_t0_t1
esttab using "F:\ test.cvs", b(2) ci(2) eform wide onecell ///
            nogaps noobs nodepvars nolines nonotes nonumbers star(+ 0.10 * 0.05 ** 0.01 *** 0.001) keep(xedu_dum) replace

The confidence interval in the output was: 1.041046 - 9.947009, whereas the esttab table displayed: [1.06,9.78]. While the p-values also differed slightly, the OR and the standard error were the same in the regression output and in the table. 

In general, the esttab table shows somewhat smaller confidence intervals than the regression output and the difference does not occur in analyses without multiple imputation. It therefore seems likely that this difference is due to the multiple imputation method. However, I always assumed that -eststo- simply saved the results of the original analyses and that analyses method would not matter.

Can anyone explain the cause of this disparity? 

Moreover, I would like to know which estimate is correct (esttab table or regression output) and whether there is an adjustment in the command that solves this problem?

Thanks and kind regards, 

Marlous Tuithof


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/