Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


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

st: how to test a published prediction model in your own data


From   Pavel Roshanov <[email protected]>
To   [email protected]
Subject   st: how to test a published prediction model in your own data
Date   Sun, 26 Jan 2014 13:31:59 -0500

Hello statalisters,

I would like to assess in my own data the predictive performance of an
existing model found in the literature. I can simply calculate a new
variable that multiplies the model coefficients (B1, B2, B3) by the
values of their respective variables (X1, X2, X3) but I'm not sure how
to estimate the uncertainty around the resulting predictions. There
appear to be several ways to calculate prediction intervals and I was
wondering if there is some shortcut in stata. As far as I can tell,
the usual postestimation -predict-  commands and options don't work in
this context.

When doing this with a logistic regression model, I want to calculate
the C statistic and somehow assess calibration through, perhaps, Brier
score or H-L GoF test.

Do you have any advice or references?

Regards,

Pavel

---
Pavel Roshanov, MSc
MD candidate, Schulich School of Medicine and Dentistry, Western University
PhD candidate, Clinical Epidemiology and Biostatistics, McMaster University
*
*   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/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index