smooth_effects: Extract fitted covariate smooths

In tulpa: Templated Unified Library for Posterior Approximation in Bayesian Hierarchical Models

Extract fitted covariate smooths

Description

The grid-marginalized posterior mean of each s(x) term's latent values, one estimate per node (bin midpoint or unique covariate value). The latent block values are read from the fit's per-grid modes, weighted by the hyperparameter grid weights.

Usage

smooth_effects(object, term = 1L)

Arguments

object	A tulpa_fit from tulpa() with s(...) term(s) in the formula.
term	Which smoother: index or covariate name. Default 1.

Value

A data frame with columns x (node location) and estimate (the posterior-mean smooth at that node), with the covariate name as an attribute "var".

See Also

field_effects() for a spatial field, ranef() for random effects.

Examples

set.seed(1)
d <- data.frame(x = runif(300, -2, 2))
d$y <- rpois(300, exp(0.3 + sin(2 * d$x)))
fit <- tulpa(y ~ s(x), data = d, family = "poisson")
sm <- smooth_effects(fit)
plot(sm$x, sm$estimate, type = "l")

tulpa documentation built on Oct. 19, 2026, 5:09 p.m.