A statistical artifact that occurs in survival analysis when the probability of including an individual observational unit in a sample is related to its survival time. The consequence is that the sample mean becomes a biased estimator of the population mean.
Family risk of disease is often assessed by using case-control studies based on referent databases. Referent databases collect family histories of cases and the cases are included in the database if one family member was diagnosed with disease.
In this sampling procedure, the probability a family is selected into the sample is related to family size. The larger the family is, the more likely to find the family in the database. Thus, in this sampling scheme the probability of sample selction is proportional to family size.
Last, J. 1983. A Dictionary of Epidemiology, New York, NY: Oxford Medical Publications.
Gordis, L. 2000. Epidemiology, 2nd ed., Philadelphia, PA: Elsevier Science.
|