modelsearch: Model searching

In kkholst/lava: Latent Variable Models

Model searching

Description

Performs Wald or score tests

Usage

modelsearch(x, k = 1, dir = "forward", type = "all", ...)

Arguments

x	lvmfit-object
k	Number of parameters to test simultaneously. For equivalence the number of additional associations to be added instead of rel.
dir	Direction to do model search. "forward" := add associations/arrows to model/graph (score tests), "backward" := remove associations/arrows from model/graph (wald test)
type	If equal to 'correlation' only consider score tests for covariance parameters. If equal to 'regression' go through direct effects only (default 'all' is to do both)
...	Additional arguments to be passed to the low level functions

Value

Matrix of test-statistics and p-values

Author(s)

Klaus K. Holst

See Also

compare, equivalence

Examples

m <- lvm();
regression(m) <- c(y1,y2,y3) ~ eta; latent(m) <- ~eta
regression(m) <- eta ~ x
m0 <- m; regression(m0) <- y2 ~ x
dd <- sim(m0,100)[,manifest(m0)]
e <- estimate(m,dd);
modelsearch(e,messages=0)
modelsearch(e,messages=0,type="cor")

kkholst/lava documentation built on Feb. 22, 2024, 4:07 p.m.