Burton P R, Palmer L J, Jacobs K, Keen K J, Olson J M, Elston R C
Genetic Epidemiology Unit, Department of Epidemiology and Public Health, University of Leicester, United Kingdom.
Am J Hum Genet. 2000 Dec;67(6):1505-14. doi: 10.1086/316899. Epub 2000 Nov 14.
It is commonly assumed that the parameter estimates of a statistical genetics model that has been adjusted for ascertainment will estimate parameters in the general population from which the ascertained subpopulation was originally drawn. We show that this is true only in certain restricted circumstances. More generally, ascertainment-adjusted parameter estimates reflect parameters in the ascertained subpopulation. In many situations, this shift in perspective is immaterial: the parameters of interest are the same in the ascertained sample and in the population from which it was drawn, and it is therefore irrelevant to which population inferences are presumed to apply. In other circumstances, however, this is not so. This has important implications, particularly for studies investigating the etiology of complex diseases.
人们通常认为，经过校正以考虑样本选择偏倚的统计遗传学模型的参数估计值，能够估计出最初从中抽取样本的总体人群中的参数。我们证明，只有在某些特定的受限情况下才是如此。更一般地说，经过校正以考虑样本选择偏倚的参数估计值反映的是样本中的参数。在许多情况下，这种视角的转变并不重要：感兴趣的参数在样本及其所抽取的总体中是相同的，因此推断适用于哪个总体并不相关。然而，在其他情况下并非如此。这具有重要意义，特别是对于研究复杂疾病病因的研究而言。
