Introducing the Psychometrics Lab

Dr Yunxiao Chen and Professor Irini Moustaki introduce the work of the LSE Department of Statistics’ Psychometrics Lab.

The 'big data' era sees every person generating a deluge of high-dimensional data, from clicks and purchases to sensor readings and survey responses. Yet the complexity of this data is often driven by unobserved human traits—core factors such as intelligence, personality, attitudes, and latent abilities. Measuring these crucial latent constructs is the foundational work of psychometrics. The Psychometrics Lab at the LSE Department of Statistics is dedicated to pushing the boundaries of this field with both rigour and creative innovation.

What is psychometrics?

Psychometrics is the discipline concerned with measurement in social and behavioural sciences. It addresses questions such as: how can we design a test or questionnaire that reliably distinguishes different levels of ability or trait? How can we ensure that the measurement is fair across groups to allow us to make group comparisons? How can we combine multiple test items to infer an underlying latent trait? What is the optimal number of latent factors required to explain a set of human behaviours, and what do these factors represent?

In practical settings, psychometric methods underpin educational testing, such as standardised exams; psychological scales such as personality inventories and mental health instruments; and increasingly ‘gamified’ assessments or adaptive testing formats. Employers and institutions can use psychometric tests to assess aptitudes, reasoning skills, personality traits, and emotional or behavioural tendencies.

A key challenge in this discipline is ensuring that such instruments are: ‘valid’ (that they measure what we intend; ‘reliable’ (that measurement is consistent and not overly noisy); and ‘fair’ (that they work equivalently for different populations). Modern psychometrics also draws on computational algorithms, high-dimensional statistical inference tools, machine learning, and latent variable modelling to provide a rich and flexible framework that meets its aims.

The Psychometrics Lab

Established in 2019, the Psychometrics Lab is embedded in LSE’s Department of Statistics. It brings together faculty members, postdoctoral researchers, PhD students, and visiting scholars who share interests in modern psychometric methodology.

We have led the Lab since its inception, and our research spans latent variable models and their computation, categorical data analysis, unsupervised machine learning, statistical theory for psychometrics, adaptive measurement with sequentially observed data, missing data analysis, and applications to problems in education, psychology, marketing, sociology, and political science. We publish in top-tier journals in psychometrics, statistics/social statistics, and machine learning, such as Psychometrika, British Journal of Mathematical and Statistical Psychology, Journal of the American Statistical Association, Biometrika, Journal of the Royal Statistical Society, Series A (Statistics in Society), and Journal of Machine Learning Research.

A forthcoming book, ‘Analysis of Multivariate Social Science Data: Statistical Machine Learning Methods’, co-authored by Moustaki, Chen, Steele and Bartholomew, is due for publication in February 2026. The third edition covers cutting-edge methods and tools for analysing multivariate data. It expands its topics to include graphical modelling, models for longitudinal data, structural equation models for categorical variables, and latent class analysis for ordinal, nominal, and continuous variables. The book also connects the topics to the terminology and principles of machine learning, intended to help readers grasp the links between multivariate analysis methods and advancements in data science. It comes with a website that provides data and code for replicating the analysis of all the book examples.

Training the statisticians and psychometricians of tomorrow, and building capacity, are top priorities. We are incredibly proud that the Lab is functioning as a hub for graduate students, visiting students and postdocs to acquire technical skills, combine domain knowledge with measurement tools, and then translate that into applied research. Because psychometric models are applicable across many social and behavioural disciplines, the Lab is well-positioned for cross-cutting collaborations. We are especially interested in hearing from prospective PhD candidates with a strong quantitative background and an interest in our core topics.

As the Lab evolves, our focus will continue to be on pushing the boundaries of measurement in complex data settings, developing robust and scalable algorithms, and promoting methodological rigour across applied fields. Specifically, we see the future of psychometrics as an integration of measurement and technology, enabling more accurate and dynamic measurement of human traits and further enabling the measurement to provide people with better personalised experiences. To this end, the Lab works on research projects that combine traditional psychometric theory and methods with cutting-edge techniques, including advanced latent variable models, stochastic optimisation, reinforcement learning, and representation learning, to build psychometric tools for the future. We stay at the frontiers of research in Psychometrics, embracing the new technologies and the transformation they are already bringing to the discipline.

Discover more about our group and research on the Psychometrics Lab website: https://psychometricslab.com/

By Dr Yunxiao Chen and Professor Irini Moustaki, Deputy Head of Department (Education), LSE Department of Statistics.