WITHDRAWN: Measures of Association for Longitudinal Data: Applications to HIV/AIDS

• Author(s): Sujay Datta+ and Hexin Zhang and Li Qin and Steven Self
• Keywords: Dependency ; Dynamical Correlation ; Longitudinal data

Abstract

Measuring association between two variables and testing for independence are two of the fundamental tasks in statistical inference and there is extensive literature on how to do these in the univariate case, including Pearson's correlation, Spearman's rank correlation and Kendall's tau coefficients. In the case of multivariate, longitudinal or time-course data, however, relatively little is available for this purpose.The few that are available (e.g. dynamical correlation by Dubin and Muller (2005) and Opgen-Rhein and Strimmer (2006) or odds ratio for correlated binary data by Lipsitz et al. (1991)) are either computationally challenging or dependent on specific assumptions about the underlying distribution or data-type. Here we introduce a number of association measures for longitudinal data (continuous, discrete numerical or ordinal) that are conceptually simple and distribution-free.