The bamlss package provides a modular computational framework for distributional regression models (and beyond). Users can specify complex models using the familiar formula syntax of R. The package supports commonly used extractor functions like plot(), summary(), predict(), etc. Moreover, users can specify their own model term constructors or even model fitting engines, Bayesian or frequentist.

Flexible model terms


Spatial effects


Prediction


Model diagnostics


Extensible model terms and model fitting engines

## Posterior mode optimization.
bfit <- function(x, y, family, start = NULL, weights = NULL, offset = NULL,
  update = "iwls", criterion = c("AICc", "BIC", "AIC"), ..) { }

## MCMC Sampler.
GMCMC <- function(x, y, family, start = NULL, weights = NULL, offset = NULL,
  n.iter = 1200, burnin = 200, thin = 1, verbose = TRUE, step = 20, ...) { }

## Gradient boosting.
boost <- function(x, y, family, weights = NULL, offset = NULL,
  nu = 0.1, nu.adapt = TRUE, df = 4, maxit = 400, ...) { }

## Lasso.
lasso <- function(x, y, start = NULL, adaptive = TRUE,
  lower = 0.001, upper = 1000,  nlambda = 100, lambda = NULL, ...) { }