Skip to content
Prev 18336 / 20628 Next

Precision about the glmer model for Bernoulli variables

Thanks for all this details about correlation.

Just to be clear about the nomenclature: when a model is called
??A-B??, A is for the cumulative distribution function used as link
between pi and the linear predictor including the random effect(s) or
its reciprocal, and B is for the distribution of random effect(s), is
this right?

If this is right, for the ? double binomial ?, it would mean that the
link function is a step function, and not a monotonous strictly
increasing function. Since it would then not be a bijection between R
and [0,1], doesn't it introduce indetermination and strong constraints
on the possible values of pi? I guess I'm wrongly interpreting what
means ? double binomial ?.

Sorry for this naive questions, I'm discovering the field...

Best regards,
On Tue, Apr 21, 2020 at 07:59:48AM +0000, John Maindonald wrote: