The use of a statistical test with low power in medical research is akin to the use of a diagnostic test with low sensitivity in clinical practice: Even when the treatment being evaluated is actually effective, the research working without adequate power is likely to find no evidence of this effect. This fact is often ignored by researchers who expend time and resources on studies that have only a minimal chance of success, and by journal editors who mistakenly interpret the negative results to mean that the treatment is not effective. Funding is requested for development of a micro-computer program that will employ monte Carlo simulations to clarify the concept of power. In the simulation procedure the researcher uses the computer to create populations that mirror the hypothesis and then draws numerous samples from these populations; power is defined as the proportion of such replications in which the observed effect is significant. The simulation procedure enables the researcher to actually see that when the sample size is inadequate the sample effect is not likely to be significant. The program will also feature the ability to compute power and required sample size by formula and to generate graphs of power as a function sample size.
| Borenstein, M (1994) Planning for precision in survival studies. J Clin Epidemiol 47:1277-85 |
| Borenstein, M (1994) A note on the use of confidence intervals in psychiatric research. Psychopharmacol Bull 30:235-8 |
| Borenstein, M (1994) The case for confidence intervals in controlled clinical trials. Control Clin Trials 15:411-28 |
