Algebraic descriptions of nominal multivariate discrete data (Q1275413)

It is well known that multivariate discrete data can be analyzed by means of log-linear models. This paper, however, proposed an algebraic approach. The authors defined algebraic operators that lead to a parametrization in terms of the moments and show how the operators transform the cell probabilities into new parameters that can be easily characterized. The authors also developed a complete characterization of all meaningful transformations of nominal data and show its impact on the parametrization by means of moments.