ABSTRACT
This chapter shows how to define a predictor matrix that multiplies all terms in the linear predictor, so that the actual linear predictor of the model is a linear combination of the effects in the model formula. It presents how to define linear combinations of the latent effects to compute their posterior distribution. The chapter introduces the use of models with several likelihoods. The use of the predictor matrix introduced to define models where the actual linear predictors are linear combinations of the effects in the model formula. With integrated nested Laplace approximation, linear combinations on the different latent effects can be defined and their posterior marginals estimated. Linear combinations are defined by setting the effects and the associated coefficients that they have to make the linear combination. When a model with several likelihoods is defined, it may be necessary to share some terms among the different parts of the model.
