Type I and type II errors are the two classic pitfalls in statistical analysis: finding a difference when there is none (type I) and failure to find a true difference (type II). There is, in addition, another important error that regularly appears in scientific journals. This error, the type III error,1 occurs whenever the conclusions drawn are not supported by the data presented. In recent years, type III errors have been increasing in prevalence. Some illustrations drawn from recently published articles should serve to define my point. I have deliberately omitted citation of sources because my intent is to illustrate, not embarrass.
In this study, dietary fiber intake was recorded in a group of patients with appendicitis and in age- and sex-matched controls without appendicitis. The mean fiber intake in 31 patients with appendicitis was 17.4 ± 7.7 g/d, with a range of 10.65 to 31.92 g/d.