Event Details

This talk examines the replication challenges of multidimensional analysis via factor analysis, noting jingle/jangle fallacies and stressing the need for methodological transparency across statistical software in social sciences and forensic research.

Multidimensional analysis (MDA) is a well-established method for the study of textual linguistic variation. It is a method that is often used for register analyses and has been suggested as a means of explaining authorship differences (Grieve 2023). However, the process of conducting an MDA is not straightforward and some replications of the method have failed to reproduce results from other studies (Lee 2000). Recent publications have provided ‘how-to’ guidance on conducting the MDA method across a range of software to try to clarify the methodological process (Egbert and Staples 2019). But how successful has this clarification been?

This talk outlines some of the learnings and pitfalls from an attempted replication of the MDA statistical method of choice, Factor Analysis, as part of my PhD research. This research draws on the psychological notion of jingle and jangle fallacies to explore possible misconceptions in the implementation of factor analysis, following Grieder and Steiner (2022).

More broadly, this presentation aims to demonstrate via this case study, the importance of methodological transparency and sensitivity across different statistical software used across the social sciences and in forensics research more widely.