Artikel
A few random thoughts about systematic error bias analysis, and reasons for it, in observational epidemiologic research
Suche in Medline nach
Autoren
Veröffentlicht: | 6. September 2007 |
---|
Gliederung
Text
Epidemiologic research is an exercise in measurement, so is inherently susceptible to non-random error or bias. Conventionally, epidemiologists quantify uncertainty arising only from random error, and address uncertainty from systematic error only qualitatively by way of a discussion of study limitations. This talk will illustrate the potential for errors in inference that arise from this conventional paradigm, some of which are rooted in cognitive psychology. These errors tend to favor causal conclusions over bias explanations for an observed association. The talk will also address alternatives to the conventional paradigm that quantify bias and uncertainty from systematic error, so should reduce the natural tendency toward favoring causal conclusions.