You are not currently logged in.
Access JSTOR through your library or other institution:
If You Use a Screen ReaderThis content is available through Read Online (Free) program, which relies on page scans. Since scans are not currently available to screen readers, please contact JSTOR User Support for access. We'll provide a PDF copy for your screen reader.
Effects of Misclassification in Epidemiologic Studies
Warren H. Gullen, Jacob E. Bearman and Eugene A. Johnson
Public Health Reports (1896-1970)
Vol. 83, No. 11 (Nov., 1968), pp. 914-918
Published by: Sage Publications, Inc.
Stable URL: http://www.jstor.org/stable/4593448
Page Count: 5
Since scans are not currently available to screen readers, please contact JSTOR User Support for access. We'll provide a PDF copy for your screen reader.
Preview not available
Comparisons of disease rates are frequently made. The rates observed may be affected by classification errors. Some persons will be misclassified according to disease status or according to presence of an attribute, or both. Under broad assumptions, the difference in observed sample prevalence rates is never larger than the difference in true sample prevalence rates, even if there is classification error in assigning persons to groups as well as to disease categories. Thus, even if the investigator cannot quantitate the classification error and adjust for it, or if he is ignorant of it, there may be some comfort in that the comparison is a conservative one, and classification error never results in the apparent difference being larger than the real difference. There is a problem since unless one quantitates and adjusts for the classification errors, the apparent difference in rates may be substantially less than the true difference, and the investigator may well not detect a "significant" difference that really exists. A second perhaps comforting feature is that if the percentage of each type of classification error is less than 50 percent, the apparent difference in sample rates and the true difference in sample rates are in the same direction and, hence, the correct group will have the larger apparent rate. This discussion is concerned only with classification errors and their effects on differences in sample rates. Any inference from sample rates to population rates also involves the effect of sampling variability. Therefore, the difference in observed sample rates may, because of sampling variation, be larger than the difference in true population rates.
Public Health Reports (1896-1970) © 1968 Sage Publications, Inc.