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

Re: st: stata resources available for funnel plots?

From   Neil Shephard <>
Subject   Re: st: stata resources available for funnel plots?
Date   Thu, 5 Mar 2009 11:50:04 +0000

On Thu, Mar 5, 2009 at 11:30 AM, Rosa Gini
<> wrote:
> dear statalisters,
> we are developing a project for comparison of institutional performance,
> and wish to perform the analysis suggested by david spiegelhalter in his
> papers
> D Spiegelhalter, Funnel plots for institutional comparison,  Qual Saf
> Health Care 2002;11:390-391
> David J. Spiegelhalter, Funnel plots for comparing institutional
> performance, Statistics in Medicine, Volume 24 Issue 8, Pages 1185 - 1202
> in short, the idea is to plot proportions (or standardized rates,...) in
> funnel plots against a measure of their precision in order to detect
> outliers. there are available resources to perform such analysis with MS
> Excel on some NHS websites such as
> we are looking for available resources in Stata, but we only found "funnel"
> which is meant to interact with metan which is written for more complex
> comparisons (ie metanalysis of risk ratio, odds ratios, risk differences)
> any suggestion? should we try and adapt metan?

Very interesting, I've been meaning to write some code to do this, but
have been exceptionally busy on other projects which have taken

Reading the papers, I think the use of proportions (and changes
thereof), standardised rates, and continuous responses is fairly
straight-forward and using -metan-/-funnel- is justifiable.  What
-meatn-/-funnel- don't currently do at the moment though is account
for over-dispersion which is common in comparing institutions.  This
seems to basically be done by scaling (or inflating as used on pp1194
of the second reference) the variance by a factor to widen the funnel
which seems quite straight-forward really, so I _think_ you'd probably
be able to adapt -funnel- to have an option to allow for
over-dispersion and this option should probably allow for both
additive and multiplicative over-dispersion.

I'd be very interested to know how things develop and more than happy
to help test code.



"The combination of some data and an aching desire for an answer does
not ensure that a reasonable answer can be extracted from a given body
of data." ~ John Tukey

Email -
Website -
Photos -

*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index