STATS MAKE ME CRY

Hi,

I have performed a ROC curve analysis and calculated positive likelihood ratios for various test scores (the aim is to predict mental disorders from scores on a psychological distress scale). Everything looks good except that positive likelihood ratios start to decrease near the upper end of the scale (i.e., for scores ranging from 27 to 30), Interestingly, though, after 30, the positive likelihood ratios resume their upward progression. What's going on? I can e-mail you the results if you need to see them prior to providing an answer to my question.

Thanks,