國立政治大學統計學系主講人：林一夔教授（Department of Mathematics, Statistics, & Computer Science of University of Illinois at Chicago ）
學 術 演 講
學 術 演 講
題 目：Overview of Agreement Statistics for Continuous, Binary, and Ordinal Data
時 間：民國105年5月12日 (星期四) 上午10：10
This is a general overview presentation with practical examples and without much statistical formulas. It will be based on materials presented in Lin, et al (2012), Statistical Tools for Measuring Agreement, Springer, NY. We will discuss definitions of precision/ accuracy/agreement and pitfalls of some misleading approaches. For continuous data we will start with the basic scenario of two assays/raters each with one measurement. We will consider the case of random or fixed target values for un-scaled (absolute) and scaled (relative) indices with constant or proportional error structure. For categorical data we will introduce traditional approaches with the basic scenario for un-scaled and scaled indices. We will present the convergence of approaches for categorical and continuous data. For both continuous and categorical data we will present examples of a unified approach for agreement among multiple raters each with multiple replicates. We will also present examples of a flexible and general setting where the agreement of certain cases can be compared relative to the agreement of a chosen case. Many practical examples will be presented in a wide variety of situations.