The Use of Statistics in the American Journal of Health Education from 1994 through 2003: A Content Analysis

Abstract

This study identifies the extent that knowledge of selected study designs and statistical techniques may assist readers in understanding the statistical component of articles in the American Journal of Health Education. The frequency of and trend in use of selected statistical research designs and statistical methods is reported, based on 366 research and review articles in volumes 25 to 34, covering 1994 to 2003, of the Journal. Almost 85% of the articles employed a descriptive study design. An increase in cross-sectional surveys that was offset by a decrease in review articles over the study period characterized almost all of the descriptive study designs. Analytic study designs were represented primarily by clinical trials and quasi-experiments. Their use did not significantly increase over the study period. Although descriptive statistics were present in more than 83% of the articles, many other articles relied on statistics beyond the descriptive statistics, such as statistics associated with model validation (33.1%), the chi-square test (25.4%), analysis of variance (ANOVA; 17.2%), t-test/z-test (16.9%), the Pearson correlation coefficient (15.3%), and the F-test (13.1%). Beyond these no other statistical methods stand out from the rest as being favored among the authors of the Journal. Epidemiologic statistical methods were less frequently used. The estimated annual percentage change in the percentage of studies not employing statistical methods was −12.71 (95% confidence interval: −23.23, −0.76). A significant increase in use was observed for measures of central tendency/dispersion, ANOVA, Pearson correlation coefficient, and the use of validity/reliability statistics for instrument validation.