The aim of this note is to describe the algorithm proposed to fit a multivariate variogram with an intrinsic correlation model (ICM). Since this model is a particular case of the linear model of coregionalization (LMC) for which the fitting algorithm is already implemented in Isatis, the algorithm we propose tries to adapt it to the particular ICM case in order to minimize the programming effort. The note is organized as follows. In a first part, we establish the notations and briefly recall how the usual algorithm works for the LMC. We also propose a way to fix the initial values for fitting a LMC in order to increase the convergence speed. Then we describe the adaptation made for the new algorithm in the second part.
