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

RE: st: kappa and bootstrap

From   Garry Anderson <>
Subject   RE: st: kappa and bootstrap
Date   Fri, 12 Oct 2007 22:08:07 +1000

Ronan, thank you for your suggestion.
My understanding of -kapci- is that it treats all observations as coming from one sample and selects random observations with replacement from this one sample. This is in contrast to there being 6 'clusters' or 'strata' of observations that make up a sample. The -bootstrap- command of Stata allows sampling by either 'cluster' or 'strata', I am just unsure as to which is more appropriate to the situation of say 6 medical clinics, with one observer assesing two methods of measuring a binary outcome on each patient and calculating kappa.
bootstrap k=r(kappa) , cluster(clinic) rep(1000) : kap a b
bootstrap k=r(kappa) , strata(clinic) rep(1000) : kap a b
A similar situation occurs when applying McNemar's test to compare the two paired proportions over the 6 'clusters' or 'strata'.
Best wishes, Garry


From: on behalf of Ronan Conroy
Sent: Fri 10/12/2007 20:44
Subject: Re: st: kappa and bootstrap

On 11 Oct 2007, at 23:12, Garry Anderson wrote:

> Dear Statalist,
> I am calculating kappa over six clusters, which range in sample 
> size from 30 to 150.
> I wish to adjust the 95% confidence intervals for possible lack of 
> independence of observations.
> The clusters could be patients within medical clinics.
> I was enquiring which of either the cluster option or the strata 
> option of the bootstrap command would be appropriate.

You might consider using -kapci- written by  Michael E. Reichenheim 
and downloadable from ssc.

P    Before printing, think about the environment
Ronan Conroy
Royal College of Surgeons in Ireland
Epidemiology Department,
120 St Stephen's Green, Dublin 2, Ireland
+353 (0)1 402 2431
+353 (0)87 799 97 95

*   For searches and help try:


© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index