[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: meta-analysis for diagnostic test
At 14:10 12/06/03 -0400, email@example.com wrote:
How can I perform a meta-analysis for a diagnostic test? I'd like to
I would use the -somersd- package, downloadable from SSC. If X is the
disease indicator and Y is the quantitative diagnostic test variable, then
Somers' D of Y with respect to X (D(Y|X)) is related to the ROC area A by
the summary ROC.
The -somersd- package saves the estimates and covariance matrices of
Somers' D as estimation results. Therefore, you could calculate a Somers' D
and its variance for each of the individual studies, extract the estimates
vextor and variance matrix for each Somers' D, concatenate the estimates
vectors for the individual studies to form a vector of Somers' D estimates,
create a diagonal matrix of the variances of the individual studies, use
-estimates store- to store these as the current estimation results, and
then use -lincom- to get a confidence interval for whatever weighted mean
of the individual study Somers' D values you wanted, and convert that
confidence interval to a confidence interval for the weighted mean ROC area.
A reference on Somers' D and ROC areas is my Stata Journal article on the
subject (Newson, 2002). A pre-publication draft of this can be downloaded
from my website at
which also contains manuals for the -somersd- package. It can be accessed
either using a browser or using the -net- command in Stata.
I hope this helps.
Newson R. Parameters behind "nonparametric" statistics: Kendall's tau,
Somers' D and median differences. The Stata Journal 2002; 2(1): 45-64.
Lecturer in Medical Statistics
Department of Public Health Sciences
King's College London
5th Floor, Capital House
42 Weston Street
London SE1 3QD
Tel: 020 7848 6648 International +44 20 7848 6648
Fax: 020 7848 6620 International +44 20 7848 6620
or 020 7848 6605 International +44 20 7848 6605
Opinions expressed are those of the author, not the institution.
* For searches and help try: