I am working on Bayesian methods for applying ordinal probit models to Likert-type item data. A Likert-type item is a statement which respondents respond to from an ordinal set of responses. Typically, these measure level of agreement such as “agree”, “neither agree nor disagree” and “disagree”. Generally, there are 5-7 response options provided. An ordinal probit model is an extension of linear regression models which is applicable here as the dependent variable, i.e. the item response, is ordinal. Bayesian methods offer an appealing alternative to frequentist statistics by allowing the incorporation of prior knowledge into the analysis, which is reflective of the scientific method. These methods are currently gaining traction in psychology research.My work is intended to provide an alternative to summing items and analysing Likert scale data. This is important because summing is an invalid manipulation of ordinal item data. The ordinal probit model also allows direct inference on a latent variable underlying the items which is often the true variable of interest but is not directly observable. My research aims to be accessible and motivating for psychology researchers who are not statistical experts and to facilitate the uptake of these methods.
Full list of publications
School of Psychology Studentship
(2010-2016) MMath Statistics (St Andrews University)
2016-2020 - PhD Cardiff University