Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: New version of -smileplot- on SSC

From   Roger Newson <[email protected]>
To   [email protected]
Subject   st: New version of -smileplot- on SSC
Date   Wed, 24 Jul 2002 11:16:00 +0100

Dear All

Thanks to Kit Baum, there is now a new version of the -smileplot- package on SSC. To describe and install the package, type -ssc describe smileplot- from Web-aware Stata.

The new version of the -smileplot- package represents a major overhaul. The package now contains two programs, -multproc- and -smileplot- (which calls -multproc-). -multproc- takes, as input, a variable containing multiple P-values and a user-specified uncorrected P-value threshold, and calculates a corrected P-value threshold using a choice of 11 multiple test procedure methods. -smileplot- calls -multproc- and then creates a smile plot, with data points corresponding to estimated parameters, parameter estimates on the X-axis, and P-values on a reverse log scale on the Y-axis. (So, the higher a data point is, the more statistically significant it is.) The Y-axis has reference lines indicating the uncorrected and corrected overall critical P-values. The line indicating the corrected critical P-value is called the parapet line, and represents an "upper confidence limit" for the set of null hypotheses that are true. The methods used to calculate the corrected critical P-value may control the familywise error rate (eg the Bonferroni, Holm and Hochberg procedures) or control the false discovery rate (eg the Simes and Benjamini-Yekutieli procedures). If a method controls the familywise error rate, and the uncorrected critical P-value is alpha, then we are 100*(1-alpha) percent confident that all rejected null hypotheses are false. If a method controls the false discovery rate, then we are 100*(1-alpha) percent confident that, if there are any rejected null hypotheses, then at least some of them are false. (In the limit, as the number of multiple tests tends to infinity, it may be argued that, if the FDR is controlled at alpha, then we are 100% confident that 100*(1-alpha) percent of rejected null hypotheses are false.) False discovery rate (FDR) is a fashionable idea in statistics at the moment (rightly or wrongly), and new FDR-controlling procedures come out all the time. I have done some preliminary certification of the package, testing its results against results in the literature.

Best wishes


Roger Newson
Lecturer in Medical Statistics
Department of Public Health Sciences
King's College London
5th Floor, Capital House
42 Weston Street
London SE1 3QD
United Kingdom

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
Email: [email protected]

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

* For searches and help try:

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