ClusterVCV: Calculate clustered standard errors for (generalized) linear...
In Brattle/BrattleExtras: Functions and other objects for Brattle projects

Description Usage Arguments Details Value Author(s) References Examples

This function is retained for backward compatibility, but is otherwise deprecated. See the multiwayvcov package for a general alternative or the plm package for a linear fixed effects model version.

1	ClusterVCV(model, data, cluster.var)

`model`	A model of class `lm` or `glm`
`data`	Data frame used in estimation
`cluster.var`	Character name of the clustering variable

The code implements the small sample correction of Cameron and Miller (2013).

Variance-covariance matrix

Charlie Gibbons

Cameron, A. Colin and Douglas L. Miller. 2013. "A Practitioner's Guide to Cluster-Robust Inference." Journal of Human Resources (forthcoming).

mtcars$cyl <- factor(mtcars$cyl)
reg <- lm(mpg ~ wt + cyl, data = mtcars)
vcv <- ClusterVCV(reg, mtcars, "cyl")

### To access the handy 'coeftest' function
library(lmtest)

### Not clustered
summary(reg)

### Clustered (it's interesting to note that the SEs get *smaller*)
coeftest(reg, vcv)