Re: st: Mlogit Nonresulting Large Dataset

[Richard Williams <richardwilliams.ndu@gmail.com>]

At 12:12 AM 9/9/2013, Mike Butterfield wrote:
> Hi Statalisters,
>
> I am using a subpopulation of the National Inpatient Sample, about 3million
> cases, and looking at different predictors for discharge types (0 = routine
> 1= died in hospital 2= to SNF/nursing home), controlling with Elixhauser
> comorbidities.
>
> Stata can give me results with many of my predictor variables within about
> 10 minutes.  However, when I use an ordinal predictor variable (eg hospital
> type small/medium/large), nothing happens, I've left it running for a day
> or two without any sign that data will be resulted.  Any ideas on what
> could be going wrong?
>
> I'm using a Mac with 8g of ram and stata 12.
>
> Here's the basic command function:
>
> svy, subpop(AGE if AGE>=18&AGE<=90): mlogit discharge HOSP_SIZE elix1 elix2
> elix3 elix4 ...[through elix 30]

Hard to tell without having the data, but a few thoughts:

* Does the exact same command work if you drop HOSP_SIZE? Make sure 
that HOSP_SIZE is really the problem. Also, I suppose you could see 
what happens when you only include HOSP_SIZE as a predictor variable.

* Have you done descriptives for HOSP_SIZE? Make sure it doesn't have 
some weird values, e.g. isn't coded 1, 2, 3, 99.

* You might try a crosstab of discharge by HOSP_SIZE to see if 
anything seems unusual, e.g. HOSP_SIZE is a constant or nearly so 
when discharge = 1. Also, HOSP_SIZ isn't super highly correlated with 
the elix variables, is it?

* You could add the -noisily- option to svy and see if anything seems 
to be happening. Maybe you'll get a bunch of "not concave" messages.

* sometimes adding the -difficult- option to the mlogit part of the 
command works miracles if the model is not converging.

* You say Stata 12. Do you really mean that, or do you mean Stata 
12.1? If you really are running Stata 12, do

update all

to get the latest version.

* As a sidelight, I think your subpop statement could be

svy, subpop(if AGE>=18&AGE<=90)

You might do something like

gen mysampl= AGE>=18&AGE<=90

and then run stats with mysampl just to make sure sample selection is 
going the way you want it to.


-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
HOME:   (574)289-5227
EMAIL:  Richard.A.Williams.5@ND.Edu
WWW:    http://www.nd.edu/~rwilliam

*
*   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/