Some seismic models that are close to classical models of analysis of variance are considered. They allow you to analyze and identify statistically significant variations of factors, which are important in data processing, as well as for solving inverse seismic problems. The focus is on the properties of these models, which distinguish them from the classical models. These properties are determined by the structure of observations characteristic of real seismic data, and the kinds of directions, allocated in the forming of factor models. As a result, new particularities appear in the model parameter estimation problem. In particular, the ambiguity increases, and to eliminate it, methods using truncation of observations and the formation of some additional conditions based on an analysis of the internal interaction between factors are proposed.
