vcov.femlm {FENmlm}

Extract the variance/covariance of a femlm fit

Description

This function extracts the variance-covariance of estimated parameters from a model estimated with femlm.

Usage

## S3 method for class 'femlm'
vcov(object, se = c("standard", "white", "cluster",
  "twoway", "threeway", "fourway"), cluster, dof_correction = FALSE,
  forceCovariance = FALSE, keepBounded = FALSE, ...)

Arguments

object

A femlm object. Obtained using femlm.

se

Character scalar. Which kind of standard error should be computed: “standard” (default), “White”, “cluster”, “twoway”, “threeway” or “fourway”?

cluster

A list of vectors. Used only if se="cluster", “se=twoway”, “se=threeway” or “se=fourway”. The vectors should give the cluster of each observation. Note that if the estimation was run using cluster, the standard error is automatically clustered along the cluster given in femlm. For one-way clustering, this argument can directly be a vector (instead of a list). If the estimation has been done with cluster variables, you can give a character vector of the dimensions over which to cluster the SE.

dof_correction

Logical, default is FALSE. Should there be a degree of freedom correction to the standard errors of the coefficients?

forceCovariance

(Advanced users.) Logical, default is FALSE. In the peculiar case where the obtained Hessian is not invertible (usually because of collinearity of some variables), use this option force the covariance matrix, by using a generalized inverse of the Hessian. This can be useful to spot where possible problems come from.

keepBounded

(Advanced users.) Logical, default is FALSE. If TRUE, then the bounded coefficients (if any) are treated as unrestricted coefficients and their S.E. is computed (otherwise it is not).

...

Other arguments to be passed to summary.femlm.

The computation of the VCOV matrix is first done in summary.femlm.

Value

It returns a N\times N square matrix where N is the number of variables of the fitted model. This matrix has an attribute “type” specifying how this variance/covariance matrix has been commputed (i.e. was it created using White correction, or was it clustered along a specific factor, etc).

Author(s)

Laurent Berge

See Also

femlm, summary.femlm, confint.femlm, resid.femlm, predict.femlm, getFE.

Examples

# Load trade data
data(trade)

# We estimate the effect of distance on trade (with 3 fixed-effects)
est_pois = femlm(Euros ~ log(dist_km) + log(Year) | Origin + Destination +
                 Product, trade)

# "normal" VCOV
vcov(est_pois)

# "white" VCOV
vcov(est_pois, se = "white")

# "clustered" VCOV (with respect to the Origin factor)
vcov(est_pois, se = "cluster")

# "clustered" VCOV (with respect to the Product factor)
vcov(est_pois, se = "cluster", cluster = trade$Product)
# another way to make the same request:
vcov(est_pois, se = "cluster", cluster = "Product")

# Another estimation without cluster:
est_pois_simple = femlm(Euros ~ log(dist_km) + log(Year), trade)

# We can still get the clustered VCOV,
# but we need to give the cluster-vector:
vcov(est_pois_simple, se = "cluster", cluster = trade$Product)