LIES, DAMNED LIES - OR STATISTICS

Purpose: The Journal of Hypertension has now produced statistical guidelines for authors. This review highlights some of the important issues in the guidelines. Type I and II errors: The Type I error is essentially the chance of a false positive result and is conventionally set to 5%. Equally important, but sometimes neglected, is the Type II error - the chance of a false negative result. Normality of data: Many statistical tests are only valid if the data are normally distributed. If the data are not normally distributed, non-parametric tests should be used. Multiple comparisons: Type I error rates must be corrected if multiple comparisons are made. Correlation coefficients: In general, regression coefficients are more useful than correlation coefficients. The latter may often be statistically significant but not convey any useful information. Confidence intervals: A confidence interval is a measure of the range which is likely to contain the true value of the parameter of interest; it conveys more information than a P value about the result of a study. Confidence intervals are also very useful for comparing studies.