Scientific vs. statistical inference: the problem of multiple contrasts in clinical research

The problem of type 1 error rates in clinical medicine research is discussed and a new measure of evaluating the incidences of type 1 error is presented. Three measures, including the experimentwise error rate, the error rate per experiment, and the percentage error rate are used to compute the type 1 error in a sample of clinical research. The results suggest the presence of a substantial incidence of type 1 error in the sampled research. The implications for statistical and scientific inferences are briefly discussed and some possible solutions explored.