At 10:50 AM 3/27/02 -0500, John W. Kulig wrote:

>JP wrote:
>
> > I have been unable to find an adequate explanation of exactly what is the
> > difference between an interclass and in intraclass correlation, and the
> > circumstances in which you would choose either.
> > Ian Kestin
>
>This is in answer to the second half of your question: Use the interclass
>when you are correlating two different things, such as height and weight. You
>can go through the entire data set of paired scores and it is obvious which
>number in each pair is a height (call it A) and which is a weight (B).
>Intraclass correlations are used when there you cannot do this. They are used
>(as one example) to get correlations for twins. If you had a set of IQ scores
>from twins, each pair of IQ scores is from a twin pair - but there is no
>basis for assigning one A and the other B. Each member of a twin duo could as
>easily be thrown into column A as column B.


to expand on this a bit without going into stat funk detail, think about 
some strategies one might look at when you want to examine the 
"relationship" between two columns of scores ... when we have (say) twin data

to make it very simple, say we have 4 twin pairs on IQs ...

  IQ A    IQ B

  103     109
   88       84
   128      119
   97        92

now, we could do a regular PPM r on the data we have ... r=.928

but, since there is no necessary logical reason why the twins are ordered 
across the lines as they are ... the set of data could have been (and made 
just as much sense)

103     109
   84       88
   128      119
   92        97

in this case, the r = .964

well, which is right? what if we rearranged the data in all possible IQA 
and IQB configurations ... did all the rs ... and took an average? we might 
say that this is the "typical" r you might get when correlating the two 
columns of values ... no matter which of the pair comes first or second

another conceptual way would be the following

the notion above is that the pair of values on a line ... that is, the twin 
PAIR ... should not be as different as, differences we might see DOWN ONE 
COLUMN OR THE OTHER  COLUMN IF there is something going on in terms of 
genetics. that is ... the between column variance (by rows) should be 
rather small ... compared to the WITHin column variance

in the extreme case, the two values in each of the columns might be the 
same for both columns like:

103     103
   84       84
   128      128
   97        97

here we see no BETWEEN column variability ... across the rows ... but, 
clearly, there is within column variability

now, this would not be usual of course but, does set the upper limit

thus, some comparison of between column variation (across the rows) versus 
within column variation ... can help to examine the issue of relationship 
between the columns of data

typically, scenarios like this might find the paired columns being for 
twins ... versus the paired columns being for non twin siblings ... and, in 
the analysis above, if the comparison of between column (across the rows) 
variances versus within column variances is different in the two cases ... 
then, we say that the role of genetics (perhaps) is the reason why the 
relationship between the sets of data are different

>--
>------------------------------------------------------------------
>John W. Kulig                        [EMAIL PROTECTED]
>Department of Psychology             http://oz.plymouth.edu/~kulig
>Plymouth State College               tel: (603) 535-2468
>Plymouth NH USA 03264                fax: (603) 535-2412
>------------------------------------------------------------------
>"Kane to kal� ke r�chto sto yal�."
>(Do a good deed and cast it to the sea)
>                         Ancient Greek saying
>------------------------------------------------------------------
>
>
>.
>.
>=================================================================
>Instructions for joining and leaving this list, remarks about the
>problem of INAPPROPRIATE MESSAGES, and archives are available at:
>.                  http://jse.stat.ncsu.edu/                    .
>=================================================================

Dennis Roberts, 208 Cedar Bldg., University Park PA 16802
<Emailto: [EMAIL PROTECTED]>
WWW: http://roberts.ed.psu.edu/users/droberts/drober~1.htm
AC 8148632401

.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to