This roundtable discussion will be focused on the convergence of agreement statistics for continuous, binary, and ordinal data. The granddaddy of the agreement statistic is mean squared deviation, or MSD=E(Y-X)2. We will discuss how the MSD being equivalently (in both estimation and statistical inference) scaled into the concordance correlation coefficient (CCC) for continuous data, weighted kappa for ordinal data, and kappa for binary data. We will discuss their relationship to a special form of intraclass correlation coefficient (ICC). For the un-scaled agreement statistics, MSD is proportionally related to the total deviation index (TDI) and coverage probability (CP) for normally distributed data, proportionally related to the weighted probability of crude correct classification for ordinal data, and is the probability of crude correct classification for binary data. Such convergence allows us to form a unified approach in assessing agreement for continuous, ordinal, and binary data from the model of basic paired samples to more complex models when we have multiple raters and each rater has multiple readings per sample. From the complex model we could also consider a flexible and general setting where the agreement of certain cases can be compared relative to the agreement of a chosen case.

Key question: • What is the current practice for assessing agreement for continuous and categorical data within the FDA and among various academia and industry environment.