sandcov {haplo.ccs}

Compute Sandwich Variance-Covariance Estimates

Description

'sandcov' computes sandwich variance-covariance estimates for the coefficients of a fitted model. These estimates may be used to calculate robust standard error estimates.

Usage

sandcov(model, id)

Arguments

model

a fitted model of class '"lm"' or '"glm"'.

id

the numeric vector used to identify subjects, expanded for each subject by the number of observations for that subject.

Details

For a model of class '"haplo.ccs"', the sandwich variance-covariance matrix is automatically provided as the object 'covariance', or may be extracted by 'vcov' (i.e., 'vcov.haplo.ccs'). See examples below.

Value

A named matrix for the covariance of the regression coefficients specified in 'model', calculated using the sandwich method.

Author(s)

Benjamin French and Thomas Lumley, University of Washington

References

French B, Lumley T, Monks SA, Rice KM, Hindorff LA, Reiner AP, Psaty BM. Simple estimates of haplotype relative risks in case-control data. Genetic Epidemiology 2006; 30(6):485-494.

Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika 1986; 73(1):13-22.

See Also

glm, haplo.ccs, lm

Examples

data(renin)

## Fit a model for covariate effects.

m1 <- glm(case ~ age + factor(race) + gender, family=binomial(link=logit))

## Obtain sandwich variance-covariance matrix.

id <- 1:length(case)
v1 <- sandcov(model = m1, id = id)

## Calculate robust standard error estimates.

se1 <- sqrt(diag(v1))

## Fit a model for haplotype and covariate effects.

m2 <- haplo.ccs(case ~ gender + age + factor(race) + haplo(geno))

## Obtain sandwich variance-covariance matrix by one of two methods.

v2 <- m2$covariance
v2 <- vcov(m2)

## Calculate robust standard error estimates.

se2 <- sqrt(diag(v2))