Integrating Design and Analysis in Longitudinal Studies.
Longitudinal studies often use a rather generic "panel design" in which subjects are measured at regular intervals (e.g., yearly). The design is typically arrived at without much consideration of how the data will be actually used. The assumption appears to be that the standard design will accommodate any mode of analysis. Recent developments in longitudinal data analysis vary in the extent to which they will tolerate or capitalize on the standard design. The workshop considered various designs and modes of analysis, and their articulation. A number of analytic approaches were reviewed, particular attention was paid to the distinction between time and state dependence.
Richard T. Campbell is Director, Research Methods Core, Health Research and Policy Centers, and Professor of Sociology at the University of Illinois at Chicago. He received his PhD from the University of Wisconsin-Madison in 1973. His substantive work focuses on the economics of aging and health. His major methodological interest is in the design of longitudinal studies. He contributed an article, Longitudinal Studies, to the Encyclopedia of Sociology (1992) and is a co-editor of Methodological Issues in Aging Research (Springer, 1988). He currently holds a research grant to study factors determining minority access to long-term care.
Paras D. Mehta is a Research Specialist at the Health Research and Policy Centers, University of Illinois at Chicago. He received his PhD in clinical psychology with a minor in quantitative psychology from the University of Houston in 1996. His methodological expertise includes longitudinal data analysis, structural equation modeling, multilevel modeling, and item response theory."
Mixed-Effect Models for the Analysis of Longitudinal Data
Mixed-effects regression models are particularly well suited to the analysis of longitudinal data. They can accommodate independent variables that are time invariant and/or time-varying; they allow for varying numbers of observations per individual; and they can be used to estimate individual-level parameters, which indicate how each individual is changing across time. The workshop described methods and applications suitable for continuous and categorical outcomes. The latter include mixed-effects logistic regression for dichotomous and nominal outcomes, and mixed-effects proportional odds and non-proportional odds models for ordinal outcomes. Additionally, software developed by the presenter (MIXREG, MIXOR, MIXNO) for performing these analyses was demonstrated.
Donald Hedeker is an Associate Professor of Biostatistics at the University of Illinois at Chicago. He received his PhD in Quantiative Psychology from the University of Chicago in 1989. His main expertise is in the development and use of advanced statistical methods for clustered and longitudinal data, with particular emphasis on mixed-effects models. He is the primary author of the MIXREG, MIXOR, MIXNO, and MIXPREG software programs. These programs perform mixed-effects regression modeling of normal, ordinal, nominal, and count data, respectively.
Hedeker, D., & Gibbons, R. D. (1994). A random-effects ordinal regression model for multilevel analysis. Biometrics, 50, 933-944.
Hedeker, D., & Gibbons, R. D. (1997). Application of random-effects pattern-mixture models for missing data in longitudinal studies. Psychological Methods, 2, 64-78.
Hedeker, D., & Mermelstein, R.J. (1996). Application of random-effects regression models in relapse research. Addiction, 91 (Supplement), S211-S229.
Hedeker, D., & Mermelstein, R. J. (1998). A multivariate thresholds of change model for analysis of stages of change data. Multivariate Behavioral Research, 33, 427-455.
Hedeker, D., & Rose, J.S. (2000). The natural history of smoking: A pattern-mixture random-effects regression model. In J.S. Rose, L. Chassin, C.C. Presson, S.J. Sherman, (Eds.), Multivariate Applications in Substance Use Research (pp. 79-112). Hillsdale, NJ: Lawrence Erlbaum Associates.