Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: deming regression


From   ymarchenko@stata.com (Yulia Marchenko, StataCorp)
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: deming regression
Date   Thu, 28 Feb 2008 12:47:17 -0600

Subramanian Swaminathan <ssubra@yahoo.com> asks about the input parameters to
the user-written command -deming-:

> I have a problem in giving inputs to the 'deming' command in stata. From a
> paper by "Cornbleet, PJ and Gochman, N (1979).  Incorrect least squares
> regression coefficients in method-comparison analysis, Clinical Chemistry 25,
> 432-438", I understand that they used a parameter called 'lambda', which is
> the ratio of the within-individual variance of the indepdent variable and the
> within-individual variance of the dependent variable. This parameter can be
> estimated from repeated measurements of depdent and indepdent variables.  My
> question is, do the the values of cv1 and cv2 in stata are based on the values
> of var1 and var2? if so, how is it related to the 'lambda' parameter of the
> above stated paper? If such repeat measurements not available, what value
> should be used for inputting to cv1 and cv2?

-deming-'s input parameters -cv1()- and -cv2()- correspond to the coefficients
of variations of the first and second methods (variables var1 and var2,
respectively).  These parameters reflect the precision of measurements for
each of the two methods relative to the mean values.  The coefficients of
variations are used to compute the error variance ratio "lambda" as follows:
lambda = cv2*mean(var2)/cv1*mean(var1).  Note that in -deming-'s output the
estimate of "lambda" is labeled as the "Variance ratio".

Alternatively, the precision of measurements may be obtained from the
duplicate measurements (if available) by computing the standard error of the
difference between the duplicates for each method.  The variance ratio is then
computed as a ratio of these standard errors.  This alternative specification,
however, is not available with -deming-.

In the absence of repeated measurements, Subramanian should obtain the values
of the coefficients of variations for the two methods based on previous
experience and the nature of the considered methods.  By default, -deming-
assumes coefficients of variations of 20% for each method.


--Yulia
ymarchenko@stata.com
*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index