Reliabilities of Intraindividual Variability Indicators with Autocorrelated Longitudinal Data: Implications for Longitudinal Study Designs

Intraindividual variability can be measured by the intraindividual standard deviation (ISD), intraindividual variance (ISD2), estimated hth- order autocorrelation coefficient (rho((h) over cap)), and mean square successive difference (MSSD). Unresolved issues exist in the research on reliabilities of intraindividual variability indicators: (1) previous research only studied conditions with 0 autocorrelations in the longitudinal responses; (2) the reliabilities of rho(h) and MSSD have not been studied. The current study investigates reliabilities of ISD, ISD2, rho(1), MSSD, and the intraindividual mean, with autocorrelated longitudinal data. Reliability estimates of the indicators were obtained through Monte Carlo simulations. The impact of influential factors on reliabilities of the intraindividual variability indicators is summarized, and the reliabilities are compared across the indicators. Generally, all the studied indicators of intraindividual variability were more reliable with a more reliable measurement scale and more assessments. The reliabilities of rho(1) were generally lower than those of ISD and ISD2, the reliabilities of MSSD were usually between those of ISD2 and rho(1) unless the scale reliability was large and/or the interindividual standard deviation in autocorrelation coefficients was large, and the reliabilities of the intraindividual mean were generally the highest. An R function is provided for planning longitudinal studies to ensure sufficient reliabilities of the intraindividual indicators are achieved.