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

st: Measurement error

From   Tom Simpson <>
Subject   st: Measurement error
Date   Wed, 17 Dec 2008 01:33:20 -0800 (PST)

Thanks for the advice. I will look into alpha (or just stick to experiments with 2 methods of measurement!).


----- Original Message ----
From: Nick Cox <>
Sent: Tuesday, 16 December, 2008 15:33:55
Subject: RE: st: RE: Re: Measurement error

-concord- is to be found in the archives for Stata Journal 7(3) 2007 (and shortly for Stata Journal 8(4) 2008). 

The limits of agreement are just (rough) confidence limits on a graph of difference vs mean for two methods. I don't know how a higher-dimensional analogue could be defined even if principle. I am clear that -concord- does not support it. 

You can find generalisations of concordance correlation to several measures in the literature but I haven't ever seen that they address the real scientific issues: rather, it seems, the question appealed to some methodologists. In any real situation with several measures that I have known the interest lies not in some overall measure of agreement but in looking at the fine structure of agreement or disagreement. 

There are various ideas in this vein within 

Cox, N.J. 2006. Assessing agreement of measurements and predictions in geomorphology. Geomorphology 76: 332-346                

The "geomorphology" element therein is largely immaterial, as the issues are generic. 

On a different tack, check out -alpha-.  


Tom Simpson

I was referring to the Bland and Altman's limits-of-agreement as calculated in concord (Version 3.0.7, Steichen and Cox)

*   For searches and help try:


*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index