PhD candidate in Latent Variable Models for Prediction

Are you interested in improving individualized predictions-such as determining the probability that a person will develop an anxiety disorder-by developing novel machine learning methods specialized for psychological data? Consider joining the Methodology and Statistics Unit at Leiden University as a PhD student and gain the skills and knowledge required to conduct cutting-edge research on data science and behavioural statistics as part of an interdisciplinary team.

Applications are invited for a four-year PhD position under the supervision of Dr. Zsuzsa Bakk, Dr. Julian Karch, and Prof. Mark de Rooij (Methodology and Statistics Unit). We are looking for an enthusiastic student who wants to develop their career in data science and behavioural statistics.

Project description
Individualized predictions are a critical aspect of psychology, as they enable professionals to make informed decisions in various contexts. We already mentioned the relevance of such individual predictions for clinical psychology. Another example comes from organizational psychology: When a selection psychologist or recruiter needs to choose between two candidates for a job, they must predict which one has higher future job performance based on individual data, selection instruments, and company characteristics.

Currently, mostly generic methods developed within machine learning, such as regularized regression or random forests, are employed to make predictions in psychology. This is suboptimal because these generic methods have not been designed specifically for psychological data. One issue is that they assume no measurement error, while this is almost never the case in psychology.

The field of psychometrics is concerned with creating latent variable models that are appropriate for psychological data: Latent variable models explicitly incorporate measurement error and impose restrictions on how the observed variables can be related by assuming that a set of relatively few latent (hidden) variables underlie the observed variables. From a predictive standpoint, it is highly plausible that these latent variable models lead to better predictions than the generic machine learning models currently used. This is because they allow the incorporation of expert knowledge in the form of a hypothesized model. This, in turn, should decrease the variance and subsequently lead to better predictions.

However, at the moment, it is impossible to obtain predictions for most latent variable models. This is because latent variable models were constructed within the explanatory modeling tradition, which aims to explain the relationship between variables instead of focusing on prediction.

In this project, we aim to fill this gap by developing prediction mechanisms for the most popular latent variable models. As a starting point, we will use the work of De Rooij, Karch, Bakk et al., who developed a prediction rule for structural equation models under the constraint that all observed variables are continuous.

Required qualifications:

• A MSc (completed, or to be awarded before the starting date) in a topic relevant for the PhD, such as statistics, data science, machine learning, mathematics, or artificial intelligence. Good knowledge of linear algebra, analysis, and probability theory
• Experience in R programming (or a similar language such as Matlab, Python, or Julia), Monte Carlo simulation studies, and data analysis
• A passion for interdisciplinary scientific research and psychology
• A collaborative mindset, openness to learning new skills, and willingness to share knowledge
• Attention to detail and clear communication skills
• Evidence of the ability to conduct high-quality research and write scientific text
• Good written and oral communication in English
• The ability to work independently as well as collaboratively within the research group

Desired qualification:

• Knowledge of latent variable models
• Master thesis within machine learning or latent variable models
• Familiarity with Open Science research practices and willingness to adopt them

Our organization
The Faculty of Social and Behavioural Sciences comprises four institutes: Education and Child Studies, Political Science, Psychology and Cultural Anthropology & Development Sociology. The Faculty also includes the Centre for Science and Technology Studies. The Faculty is home to 5,000 students and 600 members of staff. Our teaching and research programmes cover diverse topics varying from adoption to political behaviour. For more information, see http://www.universiteitleiden.nl/en/social-behavioural-sciences.

Terms and conditions
The appointment will initially be for 12 months and if evaluated positively extended with 3 years and must lead to the completion of a PhD thesis. The salary ranges from €2770 gross per month in the 1st year up to €3539 gross per month in the 4th year based on a fulltime employment in accordance with the Collective Labour Agreement for Dutch Universities. A part-time position is possible as well. The preferred starting date is the 1st of January 2024.

Leiden University offers an attractive benefits package with additional holiday (8%) and end-of-year bonuses (8.3 %), training and career development. Our individual choices model gives you some freedom to assemble your own set of terms and conditions. For international spouses we have set up a dual career programme. Candidates from outside the Netherlands may be eligible for a substantial tax break. More information can be found at https://www.universiteitleiden.nl/en/working-at/job-application-procedure-and-employment-conditions.

Diversity
Diversity and inclusion are core values of Leiden University. Leiden University is committed to becoming an inclusive community which enables all students and staff to feel valued and respected and to develop their full potential. Diversity in experiences and perspectives enriches our teaching and strengthens our research. High quality teaching and research is inclusive.

Applications
Please submit your application online no later than 7 November 2023 via the blue button in our application system. Applicants should submit a letter of interest accompanied by a CV and two referees' names and contacts (all online). Referees will be contacted directly by the search committee in case of shortlisting.

You can apply via the application button; no applications will be processed on the email addresses.

Enquiries from agencies are not appreciated.