Analysis of binary outcome data (for example, where outcome is defined as the portion of patients responding) is one of the primary methods used to study the efficacy of treatments. This type of analysis has wide application in the field of mental health research as well as other fields. The goal of this project is to develop a computer program that researchers can use to find the appropriate sample size for a planned study that will employ binary data as the outcome. The program will be developed as an update to a power analysis program that was developed under an earlier SBIR grant and is currently being distributed by SPSS, and will expand on the current program in four ways. (1) It will incorporate the ability to perform power analysis for a wide variety of procedures including logistic regression. (2) It will allow users to find the sample size to test for a clinically important effect as well as the null hypothesis of no effect. (3) It will allow users to find the sample size required to ensure that the effect size will be reported precisely. (4) It will allow the user to computer power for tests of clinical equivalence and bio-equivalence. Additionally, for 2x2 studies, we will develop and implement exact algorithms for equivalence testing, for tests of specific hypotheses, and for computation of tolerance intervals. This will comprise an important advance over the currently available algorithms which are based on a normal approximation.
The project will produce a micro-computer program that computes statistical power and precision for studies that employ binary outcome data. The program will have wide application in the fields of mental health research and medical research as well as the social sciences. The program will be marketed by SPSS and LEA, which have marketed our earlier programs.
