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

st: RE: harrell C statistics comparison

From   "Newson, Roger B" <[email protected]>
To   "'[email protected]'" <[email protected]>
Subject   st: RE: harrell C statistics comparison
Date   Sun, 6 Sep 2009 16:18:04 +0100

In principle, the difference between 2 Harrell's c statistics can be calculated using the -somersd- package (downloadable from SSC), together with -lincom-. As in:

somersd y x1 x2, transf(c) tdist
lincom x1-x2

which will give you a confidence interval for the difference between Hallell's c of -x1- with respect to -y- and Harrell's c of -x2- with respect to -y-. If -y- is a survival time, with a censorship status variable -cs- which is 1 for right-censored data, -1 for left-censored data and 0 for uncensored data, then you should type

somersd y x1 x2, transf(c) tdist
lincom x1-x2

This will be a valid comparison if -x1- and -x2- are 2 alternative predictive scores for y, as long as the dataset used to estimate the confidence limits for Harrell's c is not the dataset used to derive the scores in the first place. If you only have 1 dataset, then this should probably be split into a training set (to find the 2 nested models), and a test set (to calculate the 2 Harrell's c-indices and their difference).

I hope this helps.

Best wishes


Roger B Newson BSc MSc DPhil
Lecturer in Medical Statistics
Respiratory Epidemiology and Public Health Group
National Heart and Lung Institute
Imperial College London
Royal Brompton Campus
Room 33, Emmanuel Kaye Building
1B Manresa Road
London SW3 6LR
Tel: +44 (0)20 7352 8121 ext 3381
Fax: +44 (0)20 7351 8322
Email: [email protected] 
Web page:
Departmental Web page:

Opinions expressed are those of the author, not of the institution.

-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf Of Eduardo Nunez
Sent: 04 September 2009 20:41
To: [email protected]
Subject: st: harrell C statistics comparison

Dear Statalister,

A reviewer from a medical journal has asked me to provide 95% CI and
p-value for two nested Cox models of a Harrell C statistics
Please help.

For instances,
Model1= stcox x y, efron
estat concordance=harrell1
Model2 stcox x y z, efron
estat concordance=harrell2

harrell2 - harrell1......xx, 95%CI(xx-yy); p=x.xx
or maybe harrel2/harrel1 (I am not sure which way)

Maybe boostraping the difference.

I would appreciate any help on this issue.


*   For searches and help try:

*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index