Analyzing Ratings With Correlated Intrajudge Measurement Errors