Nonnegative Outcomes Assessed with Tweedie Distribution (110 Patients)

Abstract

Like the gamma regression (Chap. 30), Tweedie regression (named after Tweedie, a statistician from Liverpool (1984)), is generally better adequate for nonnormal data than the traditional linear regression. It can be used for statistical testing of nonnegative data with a continuous outcome variable and fits such data often better than does the normal frequency distribution, particularly when magnitudes of benefits or risks is the outcome, like costs. It is often used in marketing research. This chapter is to assess whether tweedie distributions are also helpful for the analysis of medical data, particularly those with outcome health and quality of life scores.