You mention Bland-Altman plots as if everyone knew what
they were. I doubt that this is so, especially on
a multidisciplinary list.
Incidentally, the British medical statisticians Martin Bland
and Douglas Altman have both been keen Stata users for many years,
although I don't think either ever signed up for Statalist.
Doug Altman came to one of the early London users' meetings
and contributed various inserts in the early years of the
STB.
They have persistently and energetically advocated plots of
difference versus mean as a way of comparing two methods,
although plots called Bland-Altman often carry other summary
information.
The same basic idea was used earlier in medical statistics
by
Oldham, P.D. 1962. A note on the analysis of
repeated measurements of the same subjects.
Journal of Chronic Diseases 15:969--977.
1968. Measurement in Medicine: Ihe Interpretation
of Numerical Data. London: English Universities Press.
(I owe these references to Anders Skrondal.) And in
statistics more generally, Tukey urged the usefulness
of such plots even earlier.
That's a long preamble to an answer to your question.
Difference vs mean, or equivalently difference vs
sum, is just a rotation (and a rescaling) of a scatter
plot. It thus rests on no more assumptions than a scatter
plot. The only question is whether it is useful
in helping people to perceive structure, or indeed
lack of structure, in bivariate data. The principle
is no more, and no less, than that any
constant difference gives you a horizontal reference
line.
Stata implementations include user-written programs
by various hands, in -concord-, -pairplot- and -baplot-.
Recently I've realised that in adding some details
to -concord-, Tom Steichen and I have unwittingly
revisited territory surveyed by Paul Seed in STB-55.
As he in turn didn't quote our prior work from STB-43 this
may look like some sordid dispute, but in each
case no more than ignorance and oversight is to blame.
Nick
[email protected]
Hom, Willard
> I'm wondering if anyone knows if it is appropriate to use the
> Bland-Altman approach (as implemented
> in STATA) for comparing the measurements from two separate
> rating mechanisms. That is, I'm checking
> whether two rating sytems that rate the same population of
> objects can be "tested" for their
> equivalence through the Bland-Altman method (plots, etc.).
>
> In my perusal of literature that mentions use of the
> Bland-Altman method, I have only seen its
> application in the comparison of two measurement devices that
> measure 'physical qualities" (as in
> medical/health research). Has anyone seen an application
> (and a defensible one) of the method in
> psychology, sociology, or economics, where there may be an
> interest in exploring the equivalence of
> two sets of ratings of the same objects/persons/constructs....
>
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
