All Times EDT
Keywords: prediction interval, partial correlation, causal estimand
In diagnostic medicine, a new measurement is sometimes considered to replace a standard measurement for a particular diagnostic purpose. For example, glycated albumin is a potential alternative to fructosamine as a marker of hyperglycemia. As another example, Bland and Altman (BMJ, 1995) assess agreement between paired values of intramural pH and PaCO2. A common motivation is that the standard measurement is invasive, but the replacement is not. For example, can a biomarker measurement based on a blood sample replace liver biopsy for diagnosing liver fibrosis? We consider statistical evaluation of putative replacement measurements based on extensions to the concepts of individual agreement, interchangeability, and prediction.