[R] How to use quasibinomial?

Henric Nilsson hennil at statisticon.se
Fri Jul 4 18:33:45 CEST 2003

I think I can now answer my own questions:

Henric Nilsson (hennil at statisticon.se) wrote*:

>1. I can't seem to get the correct p-values from anova.glm() for the F-tests when
>supplying the dispersion argument and having fitted the model using

Using family=quasibinomial does exactly what it says: It fits a quasi-likelihood
model with mean=mu and variance=mu(1-mu). Hence, supplying the dispersion argument
results in R treating this as a known value, i.e. not estimated from data, in which
case the F-test shouldn't be used and therefore returns the same p-value as the Chi-
squared test.

>2. When using summary.glm() on a glm object fitted using family=quasibinomial the
>reported tests are t-tests. Why?

This makes sense since we're estimating the dispersion parameter. Can't really
remember what confused me here...


More information about the R-help mailing list