Postestimation with Multiple Imputations

[gicanzo]

Hi,

I have a panel in which one variable is estimated with multiple imputation. To estimate regressions, I am using mi estimate with xtabond2:

mi estimate, post: xtabond2 L(0/2) etc,,,,

Now, the problem is that with mi I do not get postestimation results, even if I add the option [post]. Particularly, I am interested in the results of the Hansen/Sargan test and AR(1)-AR(2) tests that are tipically included with xtabond2.
I have read through the mi manual and understood that it is particularly troublesome to make such postestimation with multiple imputation, but is there any way I can still get these postestimation results?
Thanks in advance for your help,

Giuseppe

[Paul T Seed]

You do not say which version of Stata you are using. I have version 11.2, and it is possible that the command differs. However, the help file is clear.

With mi estimate , post:... , you are only asking for estimated coefficients and VCE to be posted to e(b) and e(V)
To get postestimation results, you need mi estimate postestimation : ...

If this does not solve your problem, you will need to give some example data and example code that really runs to show what you are trying to do.

HTH

[gicanzo]

Thank you for your reply Paul.
I am running Stata 12.1. I followed your suggestion and ran:

mi estimate, postestimation: xtabond2 L(0/2).(ly)..........

However, I get the error "option postestimation not allowed". What else can I try?

[gicanzo]

Up. Anyone?

[daniel klein]

Pushing a thread up, without giving new information, e.g. an example Paul has asked , it not very appreciated on the list.

Paul was most likely not suggesting that you type postestimation but referred you here. Please read the help, then come back with examples or specific questions about what you have learned from the manual.

Best
Daniel

[gicanzo]

As I said, I already had a look at the manual for -mi estimate- and related ones (including -postestimation-), but did not find an answer to my question.
I had a specific question, which is: how can I obtain the Hansen/Sargan test and the AR(1)-AR(2) tests, after -mi estimate: xtabond2 ....-
The postestimation commands available after -mi estimate- do not inlcude the tests I am looking for, which are included in -xtabond2-.
I do not understand exactly what else do you mean by giving an example? I have a panel of countries with several variables, with only one of these variables being determined by multiple imputation.

[daniel klein]

xtabond2 is a user-written command (probably) from SSC, as you are asked to explain per FAQ.

You may want to consider another user-written command, mim (SSC), designed to combine estimation results from virtually any command. Whether you get the desired output, I cannot tell, and whether the output you get (if any) is appropriate or meaningful in any sense, I cannot tell either as I have no clue about this specific test. However, mim might be your best shot from a purely technical point of view.

Please note our preference for full real names (see FAQs) on Statalist.

Best
Daniel