Prof. Eric-Jan Wagenmakers & Dr Maarten Marsman
Bayesian Inference for Ordinal Data in Psychology
Many statistical methods do not respect the ordinal scale that is typical of measurement in psychology; even when they do, classical hypothesis tests have several drawbacks. Bayesian inference offers a promising alternative framework but has not often been applied to ordinal measurements. The current project aims to harmonize these traditionally disparate fields of statistical inquiry. We propose to model test statistics and use parametric yoking in order to obtain a complete Bayesian inference framework for five nonparametric tests. This framework allows researchers to quantify evidence in favor of the null hypothesis or in favor of the alternative hypothesis, and monitor such evidence continually, as the data accumulate. This flexible method of evidence monitoring is both ethical and efficient. The proposed tests will be incorporated in R and in JASP, an open source GUI for Bayesian analyses. In sum, we propose to bring together the advantages of Bayesian inference and ordinal data analysis, and disseminate these techniques among a wide audience.
NWO Graduate Programme
1 September 2015 – 1 March 2020