Re: st: Using -estout- with multiply-imputed data: how to report r2 and r2_a

[ymarchenko@stata.com (Yulia Marchenko, StataCorp)]

"Nic" <3817629@hotmail.ca> has a follow-up question about -estout- and 
-mi estimate-:

> You mentioned as a side note in your reply to use caution with the -post-
> option. In particular, you mention that the (correctly computed) p-values
> reported by -mi estimate- (and thus your wrapper, -mibeta-) may not
> correspond to the results presented when creating a table with the -estout-
> package. Or in other words, if I understand correctly, -estout- likely
> computes p-values using incorrect (not coefficient-specific) degrees of
> freedom. The result will be tables that have incorrectly reported
> statistical significance.
>
> So if I understand correctly, it sounds like -mi estimate- and -estout-
> independently compute p-values. Is this correct? I was under the impression
> that -estout- was just using the p-values already computed by -mi estimate-.
> It sounds like I should not be using -estout- to report p-values for
> multiply-imputed data.

Stata's estimation commands, including -mi estimate-, do not store p-values or
values of test statistics in estimation results.  These values are recomputed
each time the coefficient table is displayed.  The user-written command
-estout- also recomputes these values.

The routine -mi estimate- uses for displaying results has special logic for
properly computing significance levels and confidence intervals using
multiple-imputation degrees of freedom.  -estout- does not have this logic
because it was written before -mi- became available in Stata.


-- Yulia
ymarchenko@stata.com
*
*   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/