At various points in the documentation, it looks like the scale of $\phi$ is described as if its linear predictor is transformed with the logit link function. For example, see below. However, isn't $\phi$ modelled with the log link while $\mu$ uses the logit link?
At various points in the documentation, it looks like the scale of $\phi$ is described as if its linear predictor is transformed with the logit link function. For example, see below. However, isn't $\phi$ modelled with the log link while $\mu$ uses the logit link?