In this era of increasing emphasis on cost comparisons and cost- effectiveness analysis in health care delivery systems, there has been an abundance of research that relies on cost data. Since the distribution of cost data is often skewed, and it is often assumed log- normal, it is important that health care policies are based on appropriate statistical analyses of cost data. The commonly used methods for comparing average costs of two dependent samples are the paired Student's { it t-} test on the original scales, the paired Student's { it t-} test on the log-scales, and the nonparametric Wilcoxon signed rank test. However, all of these methods my not be appropriate for comparing average costs of two dependent samples. Thus, health care policies may be formulated, based on inappropriate statistical analyses of cost data. This project will develop both parametric and semi-parametric approaches that are appropriate for comparing average costs of two dependent (paired) samples. Only then can health care professionals make informed health care choices
