lavaSearch2 (version 2.0.3)

moments2: Compute Key Quantities of a Latent Variable Model

Description

Compute conditional mean, conditional variance, their first and second derivative regarding model parameters, as well as various derivatives of the log-likelihood.

Usage

moments2(
  object,
  param,
  data,
  weights,
  Omega,
  Psi,
  initialize,
  usefit,
  update.dmoment,
  update.d2moment,
  score,
  information,
  hessian,
  vcov,
  dVcov,
  dVcov.robust,
  residuals,
  leverage,
  derivative
)
# S3 method for lvm
moments2(
  object,
  param = NULL,
  data = NULL,
  weights = NULL,
  Omega = NULL,
  Psi = NULL,
  initialize = TRUE,
  usefit = TRUE,
  update.dmoment = TRUE,
  update.d2moment = TRUE,
  score = TRUE,
  information = TRUE,
  hessian = TRUE,
  vcov = TRUE,
  dVcov = TRUE,
  dVcov.robust = TRUE,
  residuals = TRUE,
  leverage = TRUE,
  derivative = "analytic"
)
# S3 method for lvmfit
moments2(
  object,
  param = NULL,
  data = NULL,
  weights = NULL,
  Omega = NULL,
  Psi = NULL,
  initialize = TRUE,
  usefit = TRUE,
  update.dmoment = TRUE,
  update.d2moment = TRUE,
  score = TRUE,
  information = TRUE,
  hessian = TRUE,
  vcov = TRUE,
  dVcov = TRUE,
  dVcov.robust = TRUE,
  residuals = TRUE,
  leverage = TRUE,
  derivative = "analytic"
)

Arguments

object

a latent variable model.

param

[numeric vector] value of the model parameters if different from the estimated ones.

data

[data.frame] dataset if different from the one used to fit the model.

Psi

[matrix] Average first order bias in the residual variance. Only necessary for computing adjusted residuals.

initialize

[logical] Pre-compute quantities dependent on the data but not on the parameters values.

usefit

[logical] Compute key quantities based on the parameter values.

update.dmoment

[logical] should the first derivative of the moments be computed/updated?

update.d2moment

[logical] should the second derivative of the the moments be computed/updated?

score

[logical] should the score be output?

information

[logical] should the expected information be output?

hessian

[logical] should the hessian be output?

vcov

[logical] should the variance-covariance matrix based on the expected information be output?

dVcov

[logical] should the derivative of the variance-covariance matrix be output?

dVcov.robust

[logical] should the derivative of the robust variance-covariance matrix be output?

Details

For lvmfit objects, there are two levels of pre-computation:

• a basic one that do no involve the model coefficient (conditionalMoment.lvm).

• an advanced one that require the model coefficients (conditionalMoment.lvmfit).

Examples

m <- lvm(Y1~eta,Y2~eta,Y3~eta)
latent(m) <- ~eta

d <- lava::sim(m,1e2)
e <- estimate(m, d)

## basic pre-computation
res1 <- moments2(e, data = d, initialize = TRUE, usefit = FALSE,
                score = TRUE, information = TRUE, hessian = TRUE, vcov = TRUE,
                dVcov = TRUE, dVcov.robust = TRUE, residuals = TRUE, leverage = FALSE,
                derivative = "analytic")
res1$skeleton$param$Sigma

## full pre-computation
res2 <- moments2(e, param = coef(e), data = d, initialize = TRUE, usefit = TRUE,
                score = TRUE, information = TRUE, hessian = TRUE, vcov = TRUE,
                dVcov = TRUE, dVcov.robust = TRUE, residuals = TRUE, leverage = FALSE,
                derivative = "analytic")
res2$moment$Omega