Longitudinal studies: design and analysis.

Objectives should determine study design, and from that the analysis should follow. Multiple objectives, some longitudinal, some cross-sectional, have implications for sampling, especially at second and subsequent occasions. The nature of the hypothesis or estimate required, and the data to be collected, will determine the frequency and timing of follow-up. Sample size must allow for follow-up failure. Difficulties with analysis result from ill-defined hypotheses, especially in relation to repeated measurements on the same outcome variable. Methods of analysis are described in relation to the hypothesis or the estimate required.