Sample selection for the National Children's Study: form must follow function.

摘要

Belanger et al. make a strong case that some degree of probability sampling is desirable and feasible for the National Children's Study and that working through health care providers offers such a mechanism. Olsen counters that probability sampling is neither necessary nor worth the incremental cost for aetiological research, needed primarily to estimate disease burden. Both are absolutely right and both points were first made nearly a decade ago in the relatively early stages of planning the National Children's Study. I chaired a panel with the mission of advising the National Children's Study planners on the design and both these arguments were made, clearly and forcefully.