09-Sel的ectionBias.pptVIP

* Preliminary analysis of similarity among 22,000 student pairs. * Science emphasizes systematic, repeatable, carefully-conducted observation. By designing and performing experiments, a scientist attempts to control all aspects of a situation in order to observe the effect of varying only those elements s/he is trying to investigate. My brother-in-law is a physicist at the Harvard-Smithsonian Center for Astrophysics in Cambridge, Massachusetts. He has been conducting an experiment to use laser interferometry to verify the law of falling objects to a higher degree of precision than has heretofore has been obtainable. A great deal of the experimental set-up consists in insulating highly-sensitive measurement apparatus from outside perturbations such as vibration of the building. Measurements will be made at night, since the apparatus will pick up even the gravitational force from a car parking outside the building. Epidemiology and other human sciences have nowhere near the precision of measurement that is available in the physical sciences and nowhere near the ability to control threats to validity. So recognizing and minimizing the effect of myriad sources of error is a constant theme in epidemiologic research. * Sources of systematic error in epidemiologic studies are generally considered under three headings: errors that may arise from the way that study participants are selected into the study, errors from classification and measurement procedures, and errors from comparisons and their interpretation. These sources of systematic error are in addition to sampling variability, often termed “random error”. * The more important the concept, the richer the vocabulary that evolves to express its various aspects. Given the importance of error, we would expect a wealth of terms to characterize it, and we will not be disappointed in that expectation. Here are some of the major terms, including several that we have already encountered. Accuracy is a general ter