Displaying 1 result from an estimated 1 matches for "covmodelcoeffs".
2008 Jan 24
2
testing coeficients of glm
...seems to be
important.
the R code example :
###
nObs <- 10
cl <- as.factor( sample(c(1,2,3),nObs,replace=TRUE) )
y <- rnorm(nObs)
model <- glm(y ~ cl)
b <- model$coefficients
H <- c(1,1,-1) # we want to test H0: Hb = 0
### the following code will NOT run unless we can compute covModelCoeffs
#the mean of Hb is
mu = H %*% model$coefficients
#the variance is HB is
var = H %*% covModelCoeffs %*% t(H)
p.val <- 2 * pnorm( -abs(mu), mean=0, sd=sqrt(var),lower.tail = TRUE)
how do i get the covariance matrix of the estimated parameters ?
thanks,
peter
P.S. the simple solution for this...