lava (version 1.8.0)

intercept: Fix mean parameters in 'lvm'-object

Description

Define linear constraints on intercept parameters in a lvm-object.

Usage

# S3 method for lvm
intercept(object, vars, ...) <- value

Value

A lvm-object

Arguments

object

lvm-object

...

Additional arguments

vars

character vector of variable names

value

Vector (or list) of parameter values or labels (numeric or character) or a formula defining the linear constraints (see also the regression or covariance methods).

Author

Klaus K. Holst

Details

The intercept function is used to specify linear constraints on the intercept parameters of a latent variable model. As an example we look at the multivariate regression model

$$E(Y_1|X)={\alpha }_1+{\beta }_{1}X$$ $$E(Y_2|X)={\alpha }_2+{\beta }_{2}X$$

defined by the call

m <- lvm(c(y1,y2) ~ x)

To fix ${\alpha }_1={\alpha }_2$ we call

intercept(m) <- c(y1,y2) ~ f(mu)

Fixed parameters can be reset by fixing them to NA. For instance to free the parameter restriction of $Y_1$ and at the same time fixing ${\alpha }_2=2$, we call

intercept(m, ~y1+y2) <- list(NA,2)

Calling intercept with no additional arguments will return the current intercept restrictions of the lvm-object.

See Also

covariance<-, regression<-, constrain<-, parameter<-, latent<-, cancel<-, kill<-

Examples

## A multivariate model
m <- lvm(c(y1,y2) ~ f(x1,beta)+x2)
regression(m) <- y3 ~ f(x1,beta)
intercept(m) <- y1 ~ f(mu)
intercept(m, ~y2+y3) <- list(2,"mu")
intercept(m) ## Examine intercepts of model (NA translates to free/unique paramete##r)