Anna Brown, PhD

Personal page

My main areas of expertise and research interests include modelling of preference data (ranking and paired comparison tasks, and forced-choice personality questionnaires), multidimensional Item Response Theory (MIRT) and multidimensional test information, Computerized Adaptive Testing (CAT) in non-cognitive domains, and modelling response biases and faking. Some of these topics are described in more detail below. Please get in touch if you are interested in collaborating.

My most important work to date, the development of the Thurstonian IRT model, received the "2010 Best Dissertation" award from the Psychometric Society.

Topic 1. Response biases and faking

One of my long-term interests is understanding the nature of response biases in non-cognitive assessments, and suggest effective ways of combating them. Specific topics include:

  1. The cognitive processes behind motivated misresponse (aka impression management or faking) in high stakes personality testing. Specifically, situational and personal characteristics linked to applicant ‘faking good’ on employment tests; or patient ‘faking bad’ on diagnostic tests for access to treatments; etc.
  2. Biases in assessments of other people or services, for example the halo effects in 360-degree assessments; etc.

Topic 2. CAT with multidimensional self-report items

Currently I am working on investigating best strategies for item selection in Computer Adaptive Testing (CAT) involving multidimensional items measuring non-cognitive domains. Examples include personality items measuring more than one trait such as in forced-choice questionnaires; or Patient Reported Outcome Measures (PROMs) where items typically measure some general factor (e.g. psychological distress) and a specific factor (e.g. anxiety).
Results of simulation studies are conclusive in showing the efficiency of CAT in these contexts compared to static tests. Specific features of non-cognitive tests such as possible negative correlations between traits and negative factor loadings favour different strategies for item selection compared to ability tests.