A new method for estimating model parameters for multinomial data (Q1281703)

This paper presents the method of population-parameter mapping. The method allows to circumvent a computational complexity problem in Bayesian analysis of complex multinomial models. Other important features of the new estimation method are: the direct probability assessment of model coherence, and the calculation of the correlational structure among the model parameters. The procedure can be also generalized to other data types (for example process models of Gaussian data).