On data and methods in investigations on parental-age effects. Comments on a paper by J. D. Erickson

An analysis of the NIS data used by Erickson (1977) has revealed that this data contains too few Down's syndrome cases with older fathers. Thus the data is less suited for an investigation of a possible paternal age effect. In addition, the statistical tests used by Erickson have low power, so that they are unable to detect a moderate paternal-age effect. Our analysis has demonstrated that discussions on paternal-age effects and similar problems should be based on high-quality data. The investigation is better based on a small high-quality material than on a large one with serious underascertainment. In the latter case one has to be extremely careful because the missing observations nearly always have common characteristics which may cause serious bias of relevant kinds.