The aim of the paper is to present a general approach to the statistical linearization based on information-theoretic measures of dependence. In the paper, a statistical linearization problem statement using the information-theoretic criterion under rather general conditions is proposed. It is based on a parameterized description of the system model under study combined with a corresponding method of estimation of the Shannon mutual information of the output variables of the initial system and its linearized model. Such a problem statement leads finally to a problem of the finite dimensional optimization. As a result, a constructive procedure of estimating the coefficients of the weight function of the linearized model is derived.