Some non-random views of statistical significance.

Although research in family medicine is growing rapidly, few family physicians have had experience or training in statistical methods. Statistical significance and P values are often misunderstood and frequently misapplied. "Significance" is arbitrary; the actual P value is of greater interest than a significant/not significant statement; P values do not measure the strength of an association; statistical significance is not equivalent to "actual" significance; P values are largely dependent on sample size; and data "dredging" is guaranteed to yield spurious results. Competent statistical consultation, careful study planning, and recognition of statistical pitfalls are important to anyone who does research, and knowledge of these areas is useful as well to anyone reading the medical literature.