HMET5140 ¨C Non-Parametric Methods and regression alternatives

Course content

The course gives an introduction to non-parametric statistics, starting with a repetition of the difference between the mean and the median and the influence of having data with a skewed distribution. Typical examples of such data within health economics are costs of stay or the length of stay.

The course also covers simple non-parametric tests for comparing groups of observations. Methods such as the bootstrap for constructing confidence intervals in the case of skewed data is also covered. Alternative regression methods such as log-linear regression, gamma and negative binomial regression models will be discussed. Computer exercises are used throughout the course.

Learning outcome

Knowledge

• Understand the problems associated with having skewed data:
• In terms of using standard methods for hypothesis testing
• In terms of constructing confidence intervals for the mean
• Learn simple methods for transforming the data, such that standard methods may still be applied in the analysis
• Understand the concept of bootstrapping, and why this is a useful method in health economics
• Understand the concept of generalized linear models, and their usefulness when analyzing skewed data

Keywords

• Non-parametric tests for skewed continuous data
• Transformation of variables
• The bootstrap - A method for constructing valid confidence intervals in the case of skewed observations
• Alternatives to standard linear regres