Adds one native temporal covariance source. temporal_indep() fits an
AR1 or OU process for each trait, temporal_dep() fits that process with an
unstructured trait covariance, and temporal_latent() fits rank-one trait
loadings. With unique = TRUE, the temporal diagonal Psi is also correlated
across occasions; it is not independent occasion noise. Temporal sources can
be added to ordinary unit and unit_obs terms. The initial cross-source
cells are replicated AR1 temporal_indep() plus one labelled kernel_indep(),
a fixed labelled phylo_indep() term, or a fixed labelled animal_indep()
term, or a fixed labelled spatial_indep() term. The spatial cell redraws
its independent SPDE field during unconditional simulation. Other
temporal-source combinations remain unavailable. For an unreplicated,
Gaussian temporal_indep() source by itself, forecast_temporal(),
profile_temporal(), bootstrap_temporal(), and compare_temporal() have
separate bounded contracts. Generic new-data prediction, generic intervals
and profiles, automatic selection, and source-pair versions of those routes
remain unavailable.
Usage
temporal_indep(formula, time, structure = "ar1", replicate = NULL)
temporal_dep(formula, time, structure = "ar1", replicate = NULL)
temporal_latent(
formula,
time,
d = 1,
structure = "ar1",
replicate = NULL,
unique = FALSE
)Arguments
- formula
A bar expression such as
0 + trait | series.- time
Bare column naming ordered occasions. AR1 requires integers and preserves their gaps; OU accepts elapsed numeric time without rescaling.
- structure
Either
"ar1"or"ou".- replicate
Optional bare column distinguishing repeated measurements at a series–occasion–trait cell.
- d
Latent rank. This version supports
1.- unique
For
temporal_latent(), include a trait-diagonal temporal Psi.
Value
A formula marker consumed by gllvmTMB().