SAMPLING DISTRIBUTIONS OF P-POS AND P-NEG

In the absence of a gold standard, assessment of clinimetric propertie s of dichotomous variables should include reporting of the proportions of positive agreement (p(pos)) and negative agreement (p(neg)). For e xample, for a patient considering whether or not to undergo elective s urgery, p(pos) represents the probability that a second physician woul d concur with a recommendation to undergo surgery and p(neg) represent s the probability that a second physician would concur with a recommen dation not to undergo surgery. This article uses a conditional binomia l distribution to derive the sampling distributions of p(pos) and p(ne g) The sampling distribution can be used as a basis for confidence int ervals and hypothesis tests.