In this chapter, two different examples, both issued from our research experience in salmon ecology, are developed to introduce Hierarchical Bayesian Models (HBM), that make up the backbone of today’s Bayesian modeling. Hierarchical (also called multilevel or random effect) models assume that the dataset being analyzed consists of a hierarchy of different groups within which records look more alike than between groups. Random effects or latent variables are probabilistic objects which are introduced to capture the variability between those groups. They are considered to be a priori drawn from a probability distribution with parameters (typically mean and variance) that will adjust to the data. A small variance will express a strong resemblance between groups, a large one will mean that the groups do not look like one another. This probability structure ties together the various layers of such a multilevel construction.