A Mixed Longitudinal Data Analysis Model

Document Type

Article

Authors

Martin A. van't Hof, University of Nymegen, Nymegen, The Netherlands
Machteld J. Roede, State University of Utrecht, Utrecht, The Netherlands
Charles J. Kowalski, The University of Michigan, Ann Arbor, Michigan

Abstract

Most growth studies to date have been based on either cross-sectional or longitudinal designs despite the fact that each of these designs has well-known shortcomings: cross-sectional designs confound developmental differences with generation (cohort) differences; and longitudinal designs confound developmental differences with time-of-measurement differences. A theoretically preferable approach is the mixed-longitudinal design, but little experience with the data analytic aspects of such studies has been accumulated. We develop strategies for the analysis of data collected in accordance with a mixed-longitudinal study design and illustrate their use on several anthropometric measurements taken as part of the Nymegen Growth Study. Mixed-longitudinal designs can be viewed as a convenient combination of the cross-sectional and longitudinal approaches and, when used in conjunction with appropriate analytic techniques, mixed-longitudinal designs represent at least a partial solution to the problems of confounding alluded to above.

Recommended Citation

van't Hof, Martin A.; Roede, Machteld J.; and Kowalski, Charles J. (1977) "A Mixed Longitudinal Data Analysis Model," Human Biology: Vol. 49: Iss. 2, Article 8.
Available at: https://digitalcommons.wayne.edu/humbiol/vol49/iss2/8