04-28-2015 05:04 PM
I am trying to use the Cohen's Kappa statistic to compare the levels of agreement between four raters for four categories. However, the statistic I'm getting from the mAGREE macro is the level of agreement of all raters for each rating
Rating Kappa Prob>Z
1 0.58 >.001
2 0.62 >.001
3 0.45 >.001
4 0.78 >.001
The table that I want to end up filling out looks something like this
|Rater 1||Rater 2||Rater 3||Rater 4||Rater 5|
What I am thinking of doing is calculating Kappa statistics for all combinations of raters. But then my question is, does that count as multiple comparisons and would it increase my levels of Type I error? Thanks.
04-28-2015 10:43 PM
Computing a p value for Cohen's kappa seems rather odd in the first place. You would never expect the null hypothesis to be true. If it were even close to true, then there wouldn't even be a point to computing an agreement statistic and even a highly significant kappa may be quite low for any practical purpose.
So, doing many kappas does increase the chance of type I error, but if there is even a tiny chance of a type I error, something is very wrong.
04-29-2015 08:17 AM
"does that count as multiple comparisons and would it increase my levels of Type I error?"
Sorry.If I was right. You messed it with ANOVA . Agree test is not ANOVA , therefore there should not be multiple comparisons ,Type I error .