The follwing is a code snippet from a power simulation
program that I'm using:
estbeta<-fixef(fitmodel)
sdebeta<-sqrt(diag(vcov(fitmodel)))
for(l in 1:betasize)
{
cibeta<-estbeta[l]-sgnbeta[l]*z1score*sdebeta[l]
if(beta[l]*cibeta>0) powaprox[[l]]<-powaprox[[l]]+1
sdepower[l,iter]<-as.numeric(sdebeta[l])
}
Estbeta recovers the fixed effects from a model fitted using lmer.
Beta is defined elsewhere and is a user specified input
that relates the data generated in the simulation to an oucome.
So, it seems pretty clear that the third line from the bottom is
a clever test of whether the confidence interval traps 0. My
question is why use beta[l]*cibeta>0 rather than
estbeta[l]*cibeta>0. Is that because in the long run the model
parameter etimates tend toward the betas specified by the user?
In other words, what really matters is the standard errors, right?
______________________________________________
[email protected] mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
