Application of mixed model analysis to ordinal survey data: An example from UK university student satisfaction data (National Student Survey NSS)

Date: Wednesday 27 May
Time: 1-2pm
Venue: UEA, SCI 3.05

In this seminar Professor Diana Kornbrot from the University of Hertfordshire will consider the application of mixed model analysis to ordinal survey data.

Ordinal data in the form of choice of responses from small ordered sets such as: ‘strongly agree, agree, neutral, disagree, strongly disagree’ are ubiquitous. Sadly, both summary and inferential statistics are often not only statistically flawed but also practically misleading. This is in spite of the fact that appropriate mixed model ordinal regression methods are now readily available, not to speak of procedures such as forest plots. Using NSS data, it will be shown how the choice of summary statistics (e.g. proportion agree (the standard) versus proportion strongly agree or cumulative proportions; and statistical procedure (e.g. ANOVA versus logistic regression) substantially affects the outcome including the rank order of institutions and subjects studied. Dangers of inappropriate analyses, easily available remedies and educational implications will be discussed.

This event is free to attend and no booking is needed.