path function

Extract pathways in model graph

Extract all possible paths from one variable to another connected component in a latent variable model. In an estimated model the effect size is decomposed into direct, indirect and total effects including approximate standard errors.

## S3 method for class 'lvm'
 path(object, to = NULL, from, all=FALSE, ...)
## S3 method for class 'lvmfit'
 effects(object, to, from, ...)

Arguments

• object: Model object (lvm)
• ...: Additional arguments to be passed to the low level functions
• to: Outcome variable (string). Alternatively a formula specifying response and predictor in which case the argument from is ignored.
• from: Response variable (string), not necessarily directly affected by to.
• all: If TRUE all simple paths (in undirected graph) is returned on/off.

Returns

If object is of class lvmfit a list with the following elements is returned - idx: A list where each element defines a possible pathway via a integer vector indicating the index of the visited nodes. - V: A List of covariance matrices for each path.

• coef: A list of parameters estimates for each path - path: A list where each element defines a possible pathway via a character vector naming the visited nodes in order. - edges: Description of 'comp2'

If object is of class lvm only the path element will be returned.

The effects method returns an object of class effects.

Note

For a lvmfit-object the parameters estimates and their corresponding covariance matrix are also returned. The effects-function additionally calculates the total and indirect effects with approximate standard errors

Examples

m <- lvm(c(y1,y2,y3)~eta)
regression(m) <- y2~x1
latent(m) <- ~eta
regression(m) <- eta~x1+x2
d <- sim(m,500)
e <- estimate(m,d)

path(Model(e),y2~x1)
parents(Model(e), ~y2)
children(Model(e), ~x2)
children(Model(e), ~x2+eta)
effects(e,y2~x1)
## All simple paths (undirected)
path(m,y1~x1,all=TRUE)

See Also

children, parents

Author(s)

Klaus K. Holst