Stat Colloquium [In-Person]: Dr. Taeho Kim

Title: Prediction via Maximum Agreement

Abstract: While Karl Pearson’s correlation coefficient (PCC) assesses the degree of linear association between two variables or paired values of two quantitative features, there exists another measure known as Lin’s Concordance Correlation Coefficient (CCC). The CCC quantifies the degree of agreement between two variables or paired values of two quantitative features. One of the fundamental challenges in statistics, machine learning, and many scientific disciplines is the problem of prediction. This problem involves utilizing the values of a set of features to predict the value of a specific target feature. In this presentation, we will show predictors designed to maximize agreement, as measured by the CCC, between the predictor and the predictand. We will also compare these Maximum Agreement Predictors (MAP) with the widely recognized Least-Squares Predictor (LSP), which minimizes mean-squared prediction error or maximizes the PCC between the predictor and the predictand. Throughout our discussion, we will provide finite and asymptotic properties of these predictors. To illustrate their practical application, we will employ two real datasets: one related to eye data and the other concerning body fat data.