Evaluating variability in specificity on three time scales. Plots of p-values for the chi-square test over various time scales for the five comparison models over a range of mean specificity values from 0.50 to 0.99, as well as p-values for the expectation-variance model. Top: calendar year of study. Middle: month of year. Bottom: day of week. The expectation-variance model has a p-value over 0.05 for the entire range of mean specificity values for all three timescales, so the null hypothesis of constant specificity is not rejected. All plots not shown are highly significant (p < 0.001) for non-constancy.
Wieland et al. BMC Medical Informatics and Decision Making 2007 7:15 doi:10.1186/1472-6947-7-15