This is a session of Topic 4: Statistics education at the post secondary (tertiary) level
(Friday 16th, 11:00-12:30)
Methods for ordinal data analysis
- Gillian Lancaster (United Kingdom)
AbstractThe use of questionnaires and rating scales is very common in the educational, behavioural and health sciences. Questionnaire instruments typically comprise items that are rated, for example, on a likert scale with 0 indicating no problem and 4 serious problems. In medical research Health Related Quality of Life is of particular interest, and many multidimensional instruments have been designed to integrate a broad range of outcomes, for example, physical functioning, psychological well-being and social functioning.
These types of outcome, measured on an ordinal scale, require methods of ordinal data analysis. To overcome this level of complexity, ordinal data may sometimes be dichotomized into a binary variable with 0 indicating no problem and 1 any type of problem, but this may result in the loss of a rich source of data about the spectrum of severity of the problem and statistical power. These challenges for the researcher are compounded when measurements are taken over time, which requires the modelling of ordinal recurrent events.
There are many reasons why methods of ordinal data analysis methods are underutilized; for example: unfamiliarity with software, requirement to implement and use specialist software, confusion about analysis strategy, model assumptions and selection, and problems with interpretation of the findings. In this session we will hear from a range of experts who have been teaching ordinal data analysis to researchers with varying levels of statistical proficiency for many years.
|Paper||Title||Presenter(s) / Author(s)|
|4C1||Teaching: a way of implementing novel statistical methods for ordinal data to researchers||Elisabeth Svensson (Sweden)|
|4C2||Fitting transition models to longitudinal ordinal response data using available software||Mojtaba Ganjali (Iran)|
|4C3||An illustration of multilevel models for ordinal response data||Ann A O’Connell (United States)|