lavaSearch2 (version 2.0.3)

coefType: Extract the Type of Each Coefficient

Description

Extract the type of each coefficient of a lvm object.

Usage

coefType(object, as.lava, ...)
# S3 method for lvm
coefType(object, as.lava = TRUE, data = NULL, ...)
# S3 method for lvmfit
coefType(object, as.lava = TRUE, ...)
# S3 method for multigroup
coefType(object, as.lava = TRUE, ...)

Value

coefType returns a data.frame when as.lava=FALSE:

• name: name of the link

• Y: outcome variable

• X: regression variable in the design matrix (could be a transformation of the original variables, e.g. dichotomization).

• data: original variable

• type: type of link

• value: if TRUE, the value of the link is set and not estimated.

• marginal: if TRUE, the value of the link does not impact the estimation.

• detail: a more detailed description of the type of link (see the details section)

• lava: name of the coefficient in lava

When as.lava=TRUE, coefType returns a named vector containing the type of each coefficient.

Arguments

object

a lvm or lvmfit object.

as.lava

[logical] export the type of coefficients mimicking lava:::coef.

...

arguments to be passed to lava::coef

data

[data.frame, optional] the dataset. Help to identify the categorical variables.

Details

A lvm can be written as a measurement model: $$Y_i = \nu + \Lambda \eta_i + K X_i + \epsilon_i$$ and a structural model: $$\eta_i = \alpha + B \eta_i + \Gamma X_i + \zeta_i$$ where \(\Psi\) is the variance covariance matrix of the residuals \(\zeta\)
and \(\Sigma\) is the variance covariance matrix of the residuals \(\epsilon\).

coefType either returns the Latin/Greek letter corresponding to the coefficients or it groups them:

• intercept: \(\nu\) and \(\alpha\).

• regression: \(\Lambda\), \(K\), \(B\), and \(\Gamma\).

• covariance: extra-diagonal terms of \(\Sigma\) and \(\Psi\).

• variance: diagonal of \(\Sigma\) and \(\Psi\).

A link denotes a relationship between two variables. The coefficient are used to represent the strength of the association between two variable, i.e. the strength of a link. A coefficient may corresponds to the strength of one or several link.

Examples

#### regression ####
m <- lvm(Y~X1+X2)
e <- estimate(m, lava::sim(m, 1e2))

coefType(m)
coefType(e)

#### LVM ####
m <- lvm()
regression(m) <- c(y1,y2,y3)~u
regression(m) <- u~x1+x2
latent(m) <- ~u
covariance(m) <- y1~y2

m.Sim <- m
categorical(m.Sim, labels = c("a","b","c")) <- ~x2
e <- estimate(m, lava::sim(m.Sim, 1e2))

coefType(m)
coefType(e)

## additional categorical variables 
categorical(m, labels = as.character(1:3)) <- "X1"

coefType(m, as.lava = FALSE)

#### LVM with constrains ####
m <- lvm(c(Y1~0+1*eta1,Y2~0+1*eta1,Y3~0+1*eta1,
          Z1~0+1*eta2,Z2~0+1*eta2,Z3~0+1*eta2))
latent(m) <- ~eta1 + eta2
e <- estimate(m, lava::sim(m,1e2))

coefType(m)
coefType(e)

#### multigroup ####
m <- lvm(Y~X1+X2)
eG <- estimate(list(m,m), list(lava::sim(m, 1e2), lava::sim(m, 1e2)))
coefType(eG)