iid2: Extract corrected i.i.d. decomposition

Description

Extract corrected i.i.d. decomposition from a gaussian linear model.

Usage

iid2(object, ...)
# S3 method for lm
iid2(object, param = NULL, data = NULL,
  bias.correct = TRUE, ...)
# S3 method for gls
iid2(object, cluster = NULL, param = NULL, data = NULL,
  bias.correct = TRUE, ...)
# S3 method for lme
iid2(object, param = NULL, data = NULL,
  bias.correct = TRUE, ...)
# S3 method for lvmfit
iid2(object, param = NULL, data = NULL,
  bias.correct = TRUE, ...)
# S3 method for lm2
iid2(object, cluster = NULL, param = NULL, data = NULL,
  robust = TRUE, ...)
# S3 method for gls2
iid2(object, cluster = NULL, param = NULL,
  data = NULL, robust = TRUE, ...)
# S3 method for lme2
iid2(object, cluster = NULL, param = NULL,
  data = NULL, robust = TRUE, ...)
# S3 method for lvmfit2
iid2(object, cluster = NULL, data = NULL, ...)

Arguments

object

a linear model or a latent variable model

...

arguments to be passed to sCorrect.

param

[named numeric vector] the fitted parameters.

data

[data.frame] the data set.

bias.correct

[logical] should the standard errors of the coefficients be corrected for small sample bias? Only relevant if the sCorrect function has not yet be applied to the object.

cluster

[integer vector] the grouping variable relative to which the observations are iid.

robust

[logical] if FALSE, the i.i.d. decomposition is rescaled such its the squared sum equals the model-based standard error (instead of the robust standard error).

Value

A matrix containing the 1st order influence function relative to each sample (in rows) and each model coefficient (in columns).

Details

If argument p or data is not null, then the small sample size correction is recomputed to correct the influence function.

Examples

Run this code

# NOT RUN {
n <- 5e1
p <- 3
X.name <- paste0("X",1:p)
link.lvm <- paste0("Y~",X.name)
formula.lvm <- as.formula(paste0("Y~",paste0(X.name,collapse="+")))

m <- lvm(formula.lvm)
distribution(m,~Id) <- sequence.lvm(0)
set.seed(10)
d <- sim(m,n)

## linear model
e.lm <- lm(formula.lvm,data=d)
iid.tempo <- iid2(e.lm, bias.correct = FALSE)
range(iid.tempo[,1:4]-iid(e.lm))

## latent variable model
e.lvm <- estimate(lvm(formula.lvm),data=d)
iid.tempo <- iid2(e.lvm, bias.correct = FALSE)
range(iid.tempo-iid(e.lvm))
## difference due to the use of the observed info matrix vs. the expected one.

## rescale i.i.d using model-based standard error
iid.tempo <- iid2(e.lvm, robust = FALSE, bias.correct = FALSE)
diag(crossprod(iid.tempo))-diag(vcov(e.lvm))

# }