Measurement and data quality in longitudinal research.

The importance of paying attention to scale levels is emphasized and it is pointed out that Steven's hierarchy of ratio, interval, ordinal, and nominal scales is too narrow; other important scale properties have to be considered. For instance, sometimes a carefully constructed variable on a nominal scale contains more information than a variable at a higher scale level. Direct versus indirect measurement and relative versus absolute measurement are also discussed and the effects of errors of measurement on the results are considered. It is not infrequent in a longitudinal setting to disregard sampling considerations, which can be very unfortunate. Such considerations, as well as the use of modern sampling theory, can considerably enhance the quality of a longitudinal study. Finally, a number of conclusions and recommendations are given for the carrying out of longitudinal research in relation to measurement issues.