MedStats Google Group

Re: ICC and pearson's r

2010-03-18T19:56:59Z

thanks

Re: ICC and pearson's r

2010-03-18T19:33:48Z

Wei,
Look here:
[link]
Ryan

ICC and pearson's r

2010-03-18T17:16:06Z

Hi:
I am reading McGraw and Wong's 1996 paper and get this
" the pearson r is a linearity index because it measures the degree to
which one variable y can be equated to another variable x by a linear
transformation (y=ax+b). ICC(C,1) on the other hand is an additivity
index because - for the case k =2 - it measures the degree to which

RE: {MEDSTATS} Comparisons of diagnostic imaging

2010-03-18T15:35:47Z

You would not do testing in this case.
If your sample is sufficiently small you will not reject the hypothesis of zero difference.
If you sample is sufficiently lagre you will even if the difference is of no practical importance.
The relevant reference is:
author = "JM Bland and DG Altman",
title = "Statistical methods for assessing agreement between two

Re: {MEDSTATS} Weighted multirater kappa with varying numbers of raters

2010-03-18T13:55:41Z

I had a go at this many years ago. I thought it was quite
straightforward. maybe I missed something. Anyway, I never published
anything, but used to give it in a lecture. You can find a LaTeX file
of this on
[link]
If nobody knows of a similar thing, maybe I should dust this off and

Comparisons of diagnostic imaging

2010-03-18T13:39:17Z

Good afternoon,
I'm hoping that someone can help me out on a question or two.
The trust I'm working for are interested in making comparisons in the
quality of diagnostic imaging using two imaging methods. The data type
for both methods is continuous and the response variable is to be the
difference between the two imaging methods for each patient,so the

Re: Weighted multirater kappa with varying numbers of raters

2010-03-18T13:05:15Z

Norman & Streiner observe (in Biostatistics: The Bare Essentials) that
weighted kappa using quadratic weights is equivalent to the most
common form of ICC:
ICC = Var(S) / [ Var(S) + Var(Res) ]
You can see Norman & Streiner's comment in Google Books-- search on
"weighted kappa using quadratic weights has a very general

RE: {MEDSTATS} Weighted multirater kappa with varying numbers of raters

2010-03-18T10:12:26Z

Thanks for the suggestion Christian. I had not come across the AC1 and AC2 statistics before but unfortunately, according to Blood & Spratt's paper in the first link you gave, "Objects not evaluated by all raters cannot be considered when determining AC1 or AC2 statistics". Back to the drawing board ....

Re: {MEDSTATS} Weighted multirater kappa with varying numbers of raters

2010-03-18T09:23:27Z

I am not sure but maybe Gwet's AC2?
See www2.sas.com/proceedings/forum 2007/186-2007.pdf and
[link]
(It is rather difficult to gather some information on this kind of kappa).
Regards,
Christian
Martin Roberts schrieb:

RE: Weighted multirater kappa with varying numbers of raters

2010-03-18T08:59:07Z

Hi Martin
sorry I can not help - but would like to know the answer -
Paul
Dr Paul Lambe
Research Fellow
Peninsula Medical School
C409, Portland Square
University of Plymouth
Drake Circus
Plymouth PL4 8AA
e-mail paul.la...@pms.ac.uk
______________________________ __________
Sent: 18 March 2010 08:38

Weighted multirater kappa with varying numbers of raters

2010-03-18T08:38:12Z

Dear all,
Does anyone know how to calculate weighted multirater kappa in a situation where each subject is rated on an ordinal scale by a different set (and number) of raters ? Or indeed whether such a variation of kappa exists?
Fleiss (1981, Statistical Methods for Rates and Proportions, p229) gives a formula (attributed to Landis & Koch) for an unweighted coefficient but it is not obvious to me whether it is possible to extend this formula to include weights for partial agreement. Can anyone help?

Re: {MEDSTATS} Analyzing NHANES Data Using R Statistical Software

2010-03-18T02:50:44Z

Hi Mike,
I'll bet this is covered in Korn and Graubard's Analysis of Health
Surveys book. I think sampling weights are needed to answer questions
that are not for specific subjects, i.e., are not fully conditioned
upon.
Frank

Re: {MEDSTATS} Using Factor scores for regression

2010-03-17T23:30:26Z

There is no way to analyze such data that is guaranteed to give you
what you want, but if I had to put money on one analysis as the most
likely to be the most informative, I would pick a component analysis
of all the variables -- IVs and DV -- together.

Re: {MEDSTATS} Using Factor scores for regression

2010-03-17T11:53:25Z

thejasvi tv wrote
<<<
Sir,

Thanks for your inputs.

We are currently carrying out a customer satisfaction survey for some of the services rendered to them. The questionnaire has >25 questions (to be answered on a scale of 1 to 10). We also have a "Overall sat" question which I want to use as DV and the remaining questions covering various aspects of our services as the IVs. Since there are many IVs and there might me multicollinearity between the IVs, the idea of using factor analysis came up....so that we get a few IVs based on factor loadings.

Re: {MEDSTATS} Using Factor scores for regression

2010-03-17T10:53:08Z

Sir,
Thanks for your inputs.
We are currently carrying out a customer satisfaction survey for some of the
services rendered to them. The questionnaire has >25 questions (to be
answered on a scale of 1 to 10). We also have a "Overall sat" question which
I want to use as DV and the remaining questions covering various aspects of