nsem: Non-linear structural equation models

In kkholst/lava.nlin: Non-linear latent variable models via StEM and Laplace approximation

Description

Fits non-linear SEM via 1. order Laplace Approximation

Usage

nsem(model, data, laplace.control = list(), control = list(trace = 1),
  method = c("nlminb", "BFGS", "ucminf"), vcov = TRUE, p, fast = FALSE,
  ...)

Arguments

model	model list
data	data.frame
laplace.control	Options to the Laplace Approximation algorithm
control	Options parsed on to optimizer (nlminb)
method	Optimization method
vcov	Boolean indicating whether calculation of asymptotic standard
p	(Optional) evaluate log-likelihood in p
fast	if TRUE do not spend time on returning lava object/graph errors should be disabled
...	Additional parameters parsed on to lower level functions

Details

The model arguments must be a named list: list( measure0, measure1, measure2, pred0, pred1, pred2, model)

where 'model' currently can be either 'nsem2' (2-factor model) or 'nsem3' (3-factor model). Each of the remaining elements can be either character vectors or formulas.

Alternatively, a list of lists can be used as the model argument in which case a list of data.frames also much be given as data argument. A multigroup model is then fitted.

Value

lava.nlin object (available methods: coef, logLik, score, vcov, print, ...)

Author(s)

Klaus K. Holst

Examples

## Not run: 
model1 <- list(measure1=~parc+supfc+antcin+postcin,measure2=~thS+putS+midbrainS,pred0=~age+bmi,model="nsem3")
b <- nsem(model1,data=dtrp)

## End(Not run)

kkholst/lava.nlin documentation built on May 20, 2019, 10:47 a.m.