In the Fels Longitudinal Study, as should occur in any long-term serial study, great efforts were made to ensure that the data collected were reliable and that this reliability was retained during the transfer of the data to computers. High levels of data quality can be achieved in a prospective longitudinal study but not in a retrospective study. Additionally, the hypotheses posed and the analyses made in the Fels Longitudinal Study ensured, as far as possible, that the maximum information was derived from the serial nature of the data. Aspects of data management and analysis in the study will be described in the sequence: (i) the need for accurate data, (ii) quality control, (iii) data management, (iv) interpolation, (v) derivation of variables, (vi) transformation of variables, and (vii) statistical analyses.

The need for accurate data

In a cross-sectional study, errors in the measurement of some individuals have little effect on the results of analyses unless these errors are large and common. If it is concluded that outlying values denote abnormal individuals or that large errors occurred during data collection, these data points can be excluded from cross-sectional analyses. This exclusion should be documented, and based on objective rules.

Large errors may be detected by comparing observed data with the distribution of values for the same variable in other groups of children of the same sex and age. For example, a recorded stature of 90 cm for a 6-year-old boy can be recognized as an outlier by comparison with the 5th percentile level (108.5 cm).