[Roaming through the methodology. XXVIII. More problems with longitudinal studies: dropouts, head-start effects and...

The Amsterdam Growth and Health Longitudinal Study with about 300 teenage boys and girls illustrates a number of problems that can occur in a longitudinal study. By adopting a more complex study design (multiple birth cohorts and a separate control group without longitudinal measurements) confounding factors may be found. A problem with the longitudinal design are the dropouts: by comparing their study results in the period before the dropout with the results of the 'stayers' in the same period, selective dropout may be identified. In a multiple longitudinal design with overlapping cohorts, head-tail effects may occur: the most reliable results are found in the middle area where there is the most overlap. Interperiod correlations may assist in determining the stability and the accuracy of the measurements.