Psychology | Methods
Classical hypothesis tests are performed on samples of a fixed size (e.g., based on an a priori power analysis). In sequential analysis, the data are continuously monitored and sampling can be terminated whenever the data show a compelling result. Thereby, sequential tests can reduce required sample sizes considerably. Thus, by saving resources, sequential analysis allows for more efficient hypothesis testing, which bears many benefits for psychological researchers.
Statistical hypothesis testing is a form of model comparison (and selection): Statistical models representing psychological hypotheses are formulated and compared in light of data. In the Bayesian framework, models are compared by means of their relative predictive accuracy. The Bayes factor captures how well one model predicted the data relative to another, thus accounting for model fit and complexity. I am interesting in developing Bayes factors for relevant scenarios in psychological research (e.g., Likert-scale data), combining the advantages of Bayesian and frequentist methods, and using deep-learning techniques to calculate Bayes factors for complex (e.g., hierarchical) models.