Towards better research: a discussion of some common mistakes in statistical analyses

Statistical errors are common in clinical research. This paper presents a set of hypothetical but realistic scenarios in complementary medicine research. Each scenario contains at least one mistake typical of those made in analyses. These are discussed. The general principles are that charts and summary statistics can be misleading or even meaningless if not chosen carefully; objectives of the research need to be clearly defined and related to statistical hypotheses; assumptions of tests need to be checked; discussing the practical implications of observed effects is better than merely quoting P-values and data trawling should be avoided.