Vollmer W M, Johnson L R, McCamant L E, Buist A S
Center for Health Research, Kaiser Permanente, Portland, Oregon 97215.
Stat Med. 1988 Jun;7(6):685-96. doi: 10.1002/sim.4780070607.
This paper explores the extent to which differences in longitudinal versus cross-sectional inference may be influenced by the choice of statistical models. Using lung function data on 524 working men, we first compare the goodness-of-fit and implication for longitudinal decline of a variety of cross-sectional models. We then compare the predicted longitudinal patterns from these models with those observed over a period of four years. In general, both approaches provide qualitatively, if not quantitatively, similar messages concerning the relative effects of smoking and age on lung function decline. Nonetheless, we acknowledge the existence of real selection and cohort effects. Although we recognize the utility of cross-sectional designs, we discourage quantitative comparisons between studies, especially longitudinal versus cross-sectional.
本文探讨了纵向推断与横断面推断之间的差异在多大程度上可能受到统计模型选择的影响。利用524名在职男性的肺功能数据，我们首先比较了各种横断面模型的拟合优度及其对纵向衰退的影响。然后，我们将这些模型预测的纵向模式与四年期间观察到的模式进行比较。一般来说，这两种方法在定性上（即使不是定量上）都提供了关于吸烟和年龄对肺功能衰退的相对影响的相似信息。尽管如此，我们承认存在实际的选择和队列效应。虽然我们认识到横断面设计的实用性，但我们不鼓励在研究之间进行定量比较，尤其是纵向研究与横断面研究之间的比较。
