Analysis of Longitudinal Data
Longitudinal data were introduced as a special case of repeated measures in Chapter 5. Such data arise when subjects are measured on the same variable (or, in some cases, variables), on several different occasions. Because the repeated measures arise, in this case, solely from the passing of time, there is no possibility of randomizing the "occasions," and it is this that essentially differentiates longitudinal data from other repeated measure situations arising in psychology, where the subjects are observed under different conditions, or combinations of conditions, which are usually given in different orders to different subjects. The special structure of longitudinal data makes the sphericity condition described in Chapter 5, and methods that depend on it for their validity, very difficult to justify. With longitudinal data it is very unlikely that the measurements taken close to one another in time will have the same correlation as measurements made at more widely spaced time intervals.