Bayesian spatial models in INLA
The linear predictor (latent Gaussian random field) is the model of a parameter of the likelihood distribution, transformed with the link function.
E.g. using log link for parameter μi:
ηi=ln(μi)=Xi⋅β+Zi
E.g. using logit link for parameter πi:
ηi=ln(πi1−πi)=Xi⋅β+Zi