|
Abstract:
|
Wald tests are commonly used for analysis, particularly when the regression model is a generalized linear model. When Wald tests are to be used for analysis it is important to also estimate the power during the design phase using this same test. Also, complex designs are often encountered (e.g., designs with interaction terms). A method of calculating power based on the Wald test is presented. This approach transforms random multivariate normal vector draws to calculate the power of the test statistic under the alternative hypothesis empirically. This method estimates power for more complicated composite hypotheses. Results are presented for data from normal, binomial, and Poisson distributions, which are among the most common distributions assumed in the medical literature. Results are compared to simulations and to other available methods for calculating power.
|
