Hierarchical Regression Explained | Multilevel Bayesian Models

How can we fit regression models to multiple groups without treating every group as completely independent? In this video, we explore hierarchical regression, also known as multilevel regression, in the context of Bayesian statistics. Hierarchical regression extends ordinary regression by allowing parameters such as intercepts and slopes to vary across groups while linking those group-specific parameters through shared population-level distributions. This creates a flexible model that combines group-specific information with information shared across groups. The central idea is: Each group gets its own regression model — but those models are related. Instead of estimating every group completely separately, hierarchical regression allows the group-level coefficients to come from a common population distribution. This means that groups with limited data can borrow information from groups with more data, producing partial pooling and often more stable estimates. For example, we might allow each group to have its own intercept: yᵢⱼ = αⱼ + βxᵢⱼ + εᵢⱼ while modelling the group-specific intercepts as coming from a shared distribution. We can extend the same idea to slopes, allowing both the baseline level and the relationship between predictors and outcomes to vary between groups. Hierarchical regression is widely used for data with natural group structures such as individuals, schools, hospitals, countries, regions, experiments, and repeated measurements. This video is part of my Bayesian Statistics Explained series, where we build Bayesian concepts and modelling methods step by step. Subscribe for clear and intuitive explanations of statistics, probability, Bayesian modelling, and data science. #bayesianstatistics #gibbssampling #mcmc