Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: robust standard errors, sureg, & suest


From   Woolton Lee <finished07@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: robust standard errors, sureg, & suest
Date   Thu, 27 Aug 2009 12:19:10 -0400

okay.  the unit of analysis of the dataset is a health insurance
company and I am imputing rates.  For example, the percent of a health
plan's diabetic members that recieved a specific treatment.  I was
actually thinking of running my own simulation study to assess the
statistical properties of the estimation technique I am using, but I
think I'll need a statistician to advise me.

Woolton


On Thu, Aug 27, 2009 at 10:13 AM, Austin Nichols<austinnichols@gmail.com> wrote:
> Woolton Lee<finished07@gmail.com> :
> I am just guessing that modeling covariances in the data well at the
> imputation step will be crucial, and that you will need a lot more
> imputations than usual, perhaps 100 instead of 5 or 10.  Just
> guessing, as I say.  You should run your own simulation--what is your
> data structure?  What are you imputing?
>
> On Thu, Aug 27, 2009 at 10:02 AM, Woolton Lee<finished07@gmail.com> wrote:
>> I am planning to estimate a multi equation model using multiply
>> imputed data (5 imputations).  I plan to use sureg and mim then use
>> suest to recover robust standard errors for the parameter estimates.
>> I am assuming this will work from what I've read in the help file but
>> I am not certain if there are problems to implementing this plan that
>> I haven't anticipated.  Could someone please confirm that as far as
>> implementing my idea, it will work?  Thanks.
>>
>> I believe one problem with what I've planned is that combining
>> multiply imputed estimates using sureg may not be guaranteed to yield
>> statistically valid results - there may not have been any simulation
>> studies to verify that this approach to estimation will be unbiased
>> and consistent, etc.  If there is anyone that can comment on this as
>> well please do.
>>
>
> *
> *   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/
>

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index