Usage ----- ^kapgof^ varlist, [level(real 0.05) kappa_options]
Description ----------- ^kapgof^ computes confidence intervals for the kappa statistic for two indepedent raters and a binary outcome. These confidence intervals for kappa have desirable small-sample properties. It also computes confidence intervals for agreement.
This confidence interval is not conditional on kappa being zero or a pre-specified non-null kappa, as some other procedures are.
It calls @kap@ and returns the same values as this. Although values for the standard error and z-score are returned, they are not printed. Presumably, they wouldn't be as useful as confidence intervals.
Examples ----------- .^kapgof^ astma1 astma2
computes the kappa statistic and confidence intervals
Remarks ------- Bloch & Kraemer (1989) suggest an arcsine transformation. Donner & Eliasziw (1992) compare this with the goodness-of fit approach and find that the GOF approach gives coverage levels that are closer to nominal, even in small samples down to as little as n=25. I am not aware of any implementations of this arcsine transformation in any widely available computer program.
The confidence interval given in Fleiss (1981) p.221 is for a pre-specified non-null kappa.
The confidence intervals for agreement are computed, but not printed at present. Use ^return list^ to see the results. They are simply a binomial confidence interval for the number of agreeing cells as a binomial trial out of the total number of tries, e.g. (a+b) out of N trials.
References ------- Donner, Eliasziw "A goodness-of-fit approach to inference procedures for the kappa statistic: confidence interval construction, significance-testing and sample size estimation", Statistics in Medicine (Stat Med), 1992: vol 11, pp. 1511-1519
Fleiss "The measurement of interrater agreement" in Fleiss: "Statistical methods for rates and proportions" (Wiley & Sons, New York) 1981; 212-235.
Bloch, Kraemer, "2x2 kappa coefficients: measures of agreement or association" Biometrics 1989; 45: 269-287
Author ------ Jan Brogger, jan.brogger@@med.uib.no