Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


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

st: Calibration and discrimination in logistic regression - McKelvey R^2


From   Mikkel Brabrand <mbrabrand@health.sdu.dk>
To   "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu>
Subject   st: Calibration and discrimination in logistic regression - McKelvey R^2
Date   Sat, 25 Aug 2012 09:17:49 +0200

I've developed a risk stratification score predicting short term mortality
in medical patients using logistic regression. The system was developed in a
development cohort and validated in two independent cohorts. To test the
model, I would like to calculate various markers for calibration and
discrimination. 

Calculating area under the receiver operating characteristic (ROC) curve and
Hosmer-Lemeshow goodness of fit as well as Brier score has been done, but I
have been unable to calculate McKelvey R^2 in the validation cohorts. I've
only located the command fitstat to calculate McKelvey R^2 but as far as
I've found out, it can only be used in the original regression and not in
the validations. 

I have all my data in a single dataset with the different cohorts identified
by the variable cohort, with 1 = development, 2 = validation 1, 3 =
validation 2. 

The model was developed using xi: logistic outcome var1 var2 i.var3 var4
var5 if cohort == 1
  
If I run fitstat after this command, I get data for the development cohort,
but how to calculate it for cohorts 2 and 3? fitstat if cohort == 2 does
nothing.

Please help!

Thanks in advance!

Mikkel Brabrand


*
*   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   |   Site index