Computational Challenges

This chapter discusses the computational requirements for obtaining the previously defined posterior models. Alternative mitigation strategies are suggested whenever these requirements cannot be met. The primary computational challenge is faced in large studies with a large number of observations. Assessment of the posterior model based on brute-force McMC approaches is unfeasible in such studies. Models such as the Kriging predictor, the Gaussian Markov model, the basis function model, the kernel predictor and the integrated nested Laplace approximation are presented in the notation of the book. The focus is on their computational efficiency. Approximations are recommended for large studies with a large set of observations.