Skip to contents

Builds a symbolized_model from a gllvmTMB fit. Covers the Gaussian and binomial B-tier latent-variable families: per-trait intercepts (0 + trait), the between-unit reduced-rank loading term (latent(0 + trait | unit, d = K)), and the optional per-trait unique variances (unique(0 + trait | unit)). Other families and covstructs return capability errors via capability_check().

Usage

# S3 method for class 'gllvmTMB'
symbolize(fit, symbols = NULL, units = NULL, context = NULL, ...)

Arguments

fit

A fitted statistical model object.

symbols

Optional named character vector mapping variable names to user-supplied LaTeX symbols, e.g. c(body_mass = "W_i", temperature = "T_i").

units

Optional named character vector mapping variable names to units, e.g. c(body_mass = "g", temperature = "C").

context

Optional short character description of the model, e.g. "avian body-size location-scale model".

...

Reserved for method-specific extra arguments.

Value

A symbolized_model object.

Confidence intervals

Fixed-effect estimates and Wald-style approximate CIs are pulled from summary(fit$sd_report, "fixed"). Loadings (\boldsymbol{\Lambda}_B) and the between-unit covariance (\boldsymbol{\Sigma}_B) are point estimates; parametric-bootstrap uncertainty via gllvmTMB::bootstrap_Sigma() is on the deferred list.

References

Nakagawa, S. (forthcoming). gllvmTMB: Generalized linear latent-variable models in TMB. https://itchyshin.github.io/gllvmTMB/