Annals of Internal Medicine · 1983 · 72 citations · 0 references
Natural SciencesBiostatisticsExperimental DesignStatisticsWomen's HealthHypothesis TestingSex DiscriminationStatistical AnalysisForced Expiratory Volume AssessmentMultiple RegressionSex Effects On Forced Expiratory VolumeUnivariate Analysis
Analyzing Research Data is de¬ signed as the second text in this two-volume set. Its purpose, the author states, is primarily to prevent the misuse of statistical procedures. That is a worthwhile goal; such misuse and the obfuscation that accompanies it should be prevented. Chapters include discussions of univariate distributions; testing of hy¬ potheses with nominal, ordinal, and continuous response variables; analy¬ sis of variance; evaluation of simple and multiple regression; presentation of a variety of more exotic multivariate analysis procedures; use of com¬ puter programs in statistical analy¬ sis; and the presentation of statistical data in research reports. As in the first text, the author draws examples from a variety of research areas, including the evalua¬ tion of birth control methods, detec¬ tion of sex discrimination, compari¬ son of uric acid levels in men and women, notation of the effects of anesthetics on arrhythmia in dogs, and evaluation of the relationship between age and forced expiratory volume in adults. In attempting to caution against the misuse of statistics, the author devotes considerable attention to lev¬ els of measurement. Levels of mea¬ surement are important, although the debates over whether data generated by lower-level measures can be appro¬ priately summarized by higher-level statistics have drawn attention from much more important issues of statis¬ tical methodology (Borgatta EF, Bohrnstedt GW: Level of measure¬ ment: Once over again, in Bohrnstedt G, Borgatta EF (eds): Social Measure¬ ment: Current Issues. Beverly Hills, Calif, Sage Publications Inc, 1981). For example, ignoring the structure of the errors in a statistical model may have a more pernicious impact on the drawing of substantive scien¬ tific conclusions than does the as¬ sumption that a dichotomous variable can be treated as if it were continu¬