Information sandwich variance estimator for crossed designs

Usage

xeffect.glm(glm.obj, g1, g2)

Arguments

glm.obj A generalised linear model object
g1 Identifiers for one grouping factor
g2 Identifies for other grouping factor

Description

Computes a sandwich variance estimator for a two-way crossed design, where correlations are present due to observations sharing either one of two grouping factors. One example is a multi-rater diagnostic reliability study where many raters rate many images and ratings are correlated if they share either the rater or the image. Another example is given in section 14.5 of the McCullagh and Nelder reference.

Value

a covariance matrix

Author(s)

Thomas Lumley

References

Lumley (1998) "Marginal Regression Modelling of Weakly Dependent Data" PhD thesis

McCullagh and Nelder (1989) "Generalised Linear Models" Chapman and Hall.

See Also

infjack.glm,gee

Examples