THE RATIONALE AND USE OF RIDIT ANALYSIS IN EPIDEMIOLOGIC STUDIES OF BLOOD PRESSURE1

Publisher Website
Google Scholar
Abstract
Kantor, S. and W. Winkelstein, Jr. (Division of Epidemiology, Univ. of California School of Public Health, Berkeley, Calif. 94720). The rationale and use of ridit analysis in epidemiologic studies of blood pressure. Amer. J. Epid., 1969, 90: 201–213.— The characteristic response instability of blood pressure is described in terms of the descriptive validity of measurement. Poor descriptive validity results in an uncontrolled proliferation of false positive and false negative findings. Due to the low level of descriptive validity in the Buffalo Blood Pressure Study, a level which may indeed characterize most epidemiologic studies of blood pressure, a scale transformation was developed, which by balancing increases in descriptive validity against decreases in statistical sensitivity, produced a solution yielding the maximum gain in descriptive validity for the lowest “cost” in statistical sensitivity. The resultant scale of measurement is shown to approximate a modified rank order scale, the properties of which correspond most closely to the scale of measurement underlying the ridit, a statistic developed by Bross to summarize the information contained in “borderland response variables”. Ridit analysis of blood pressure data is described and illustrated. Ridit analysis, in the Buffalo study, was shown to be equal in statistical power to the mean of the original observations in large samples, and probably superior in power in small samples. The latter finding suggests the need for a reexamination of the use of parametric statistics in blood pressure research. The theoretical relationship of descriptive validity to statistical power requires more comprehensive study.

This publication has 0 references indexed in Scilit: