Re: Inter-instrument agreement

On Dec 21, 9:19 am, AMY <andrea.me...@xxxxxxxxx> wrote:
Inter-instrument agreement is often measured using the intra class
correlation coefficient (ICC). Here, I want to know whether the
agreement between two instruments (or methods), let's call them y1 and
y2, varies with a third variable (=x) that is continuously
distributed. As an example one might be interested in knowing whether
the agreement between the self-reported number of binges in a given
time interval and the corresponding number using ecological momentary
assessment (EMA) is negatively related to the subject's BMI. If the
covariate was categorical, I could simply compute the ICC for each
category separately. In the case of a continuous covariate this is
obviously not possible unless I categorize the covariate beforehand
(at the cost of reduced power). My idea is now to compute a kind of
index, such as the difference between y2 and y1 and regress it against
x using the formula y2-y1=a+b*x. This is the same as using a multiple
regression model in which the coefficient of y1 is forced to be one:
y2=a+b*x+1*y1. Another, and maybe more useful model because it more
directly addresses the question at hand, would be: y2=a+b*x
+c*y1+d*x*y1. In this model one would primarily be interested in the
interaction coefficient d which tests whether the relationship between
y2 and y1 is moderated by x.

As I am unsure about how to proceed in this matter I would be grateful
to any comments regarding these thoughts, including references that
help me getting on.

Many thanks in advance

AMY

Not my area...but....Since the ICC is based on the repeated measures
anova, could you use a repeated measures ANCOVA??
.

Relevant Pages

  • Inter-instrument agreement
    ... Inter-instrument agreement is often measured using the intra class ... correlation coefficient (ICC). ... In the case of a continuous covariate this is ... regression model in which the coefficient of y1 is forced to be one: ...
    (sci.stat.consult)
  • Re: Inter-instrument agreement
    ... correlation coefficient (ICC). ... the agreement between the self-reported number of binges in a given ... regression model in which the coefficient of y1 is forced to be one: ... between two methods of clinical measurement. ...
    (sci.stat.consult)
  • Re: Life is Funny that Way
    ... "It's a miracle that we were able to finally file the Compliance Bid ... Bindra said they had sought time from Ehsan Mani, the ICC president, ... What is Pakistan's standing to demand that games be played in Dubai? ... If there was any agreement in force between the various parties ...
    (rec.sport.cricket)
  • Re: RSC fallout
    ... but prudence would have dictated that they consult with the ICC ... referee) was nearby, and at the very least consultation with him would ... the forfeiture decision. ... *agreement* from ...
    (rec.sport.cricket)
  • Re: Inter-instrument agreement
    ... correlation coefficient (ICC). ... regression model in which the coefficient of y1 is forced to be one: ... All three correlations have the same sign, ... Partial ICCs are obtained by replacing the marginal values on the ...
    (sci.stat.consult)