Learn PyINLA
Pick a topic to learn core concepts and how to use them in pyINLA.
How INLA Works
Understand the Integrated Nested Laplace Approximation and why it's fast for latent Gaussian models.
Neural Networks vs INLA
Detailed mathematical comparison of Frequentist NNs, Bayesian NNs, and INLA on the same regression problem.
Likelihoods
Choose a likelihood and link for your response (counts, proportions, survival...).
Link Functions
Identity, log, logit and more. How links connect linear predictors to responses.
Fixed Effects
Add one or more fixed-effect covariates to your model.
Random Effects
Add iid, group-specific, temporal and spatial components via model['random'].
The Latent Field
How fixed and random effects combine in the latent field to form the linear predictor.
Priors
Choose priors for fixed effects, hyperparameters and more; configure defaults or overrides.
Marginal Posteriors
Understand marginal distributions and use utility functions to evaluate, transform, sample, and summarize them.
Posterior Sampling
Draw samples from the joint posterior for derived quantities, predictions, and Monte Carlo integration.
Predictions & NAs
Generate predictions, handle missing values, and understand how pyINLA treats NAs.
Diagnostics
DIC, WAIC, log marginal likelihood and other tools to compare and assess models.
Creating Meshes & fmesher
Build spatial domains, boundaries and meshes for SPDE models using Python's fmesher wrapper.
Understanding SPDEs
Matern covariance via SPDE: the math, FEM matrices, precision matrix Q, and how it all connects.
Maps
Download administrative boundaries, build adjacency graphs, and plot posterior values on choropleth maps for Besag/BYM2 models.
Defaults
What pyINLA chooses for you, and how to override it.
Data Collection
Ready-to-use pipelines for collecting weather, boundaries, elevation, and other data for spatial modeling.