Botwinick J, Arenberg D
Exp Aging Res. 1976 Jan;2(1):55-61. doi: 10.1080/03610737608257977.
In several published studies of intelligence and adult age, a sequential design was used in which birth cohort and time of measurement were the independent variables in an ANOVA. The time span of these two variables were quite disparate; the cohort variable covered about 50 years whereas the time-of-measurement variable covered 7 or 14 years. It was demonstrated here that when time spans are disparate, even when all differences are attributable entirely to age changes, the F values typically will be larger for the cohort variable and errors in interpretations became likely. When cohort differences are statistically significant and time-of-measurement differences are not, the conclusion that age differences are "generational" and not due to age changes within individuals may be wrong.
在几项已发表的关于智力与成年年龄的研究中，采用了一种序列设计，其中出生队列和测量时间是方差分析中的自变量。这两个变量的时间跨度差异很大；队列变量涵盖了约50年，而测量时间变量涵盖了7年或14年。这里证明，当时间跨度不同时，即使所有差异完全归因于年龄变化，队列变量的F值通常也会更大，并且可能会出现解释错误。当队列差异具有统计学意义而测量时间差异不具有统计学意义时，关于年龄差异是“代际”的且不是由于个体内部年龄变化的结论可能是错误的。
