esttab and mi - no stored results

[dbraudt]

Dear all:

I am trying to use esttab and eststo with multiply imputed data and consequent analyses. The trouble I am having is that the results are not being stored. With the following code I receive an empty esttab output. Any help with this would be great.

Thanks in advance!


Code:
use DHS_emp_IMPUTED_stunting_01

global ind_vars = "empower abuse"

eststo clear
eststo: mi est: reg stunt $ind_vars
eststo: mi est: reg stunt $ind_vars cage
esttab



Output (after the two regression stored successfully):

[Timothy Mak]

-mi estimate- stores only e(b_mi) and e(V_mi). These are not picked up by -eststo-. You can however request that e(b) and e(V) be stored using the -post- option in -mi est-, e.g.:

eststo: mi est, post : reg stunt $ind_vars

HTH,

Tim

[dbraudt]

Thank you Tim. That was the fix I needed.

[Yulia Marchenko (StataCorp)]

I just wanted to mention that mi estimate's option post must be used with caution. Although it provides a way for making certain commands such as estout, esttab, and estimates table work after mi estimate by reposting e(b_mi) and e(V_mi) to e(b) and e(V), as Tim mentioned, it does not guarantee that all of the computations produced by those commands are applicable to MI results. For example, I see that esttab provides options to display p-values and confidence intervals. To provide correct values of these statistics for MI results, coefficient-specific degrees of freedom (as stored in e(df_mi) by mi estimate) must be used for their computation. My guess is that esttab does not implement this because it was developed before mi estimate existed.

In general, proper handling of MI results requires specialized logic available only to commands which understand MI results. This is one of the reasons why mi estimate does not automatically post MI results to e(b) and e(V).

[Ryan Sandler]

FYI, estout added support for using e(df_mi) in June 2014. See: http://repec.org/bocode/e/estout/history.html

[Cecilia Luini]

Thanks a lot for this post,

But is there a way to store and estate/estout the OR or IRR instead?

Thanks a lot in advance,
Cecilia Luini

[Andrew Musau]

Odds ratios or incidence-rate ratios are just exponentiated coefficients. So you need the -eform- option of estout.

esttab..., eform