An urn model with local reinforcement: a theoretical framework for a chi-squared goodness of fit test with a big sample

Motivated by recent studies of big samples, this work aims at constructing a parametric model which is characterized by the following features: (i) a "local" reinforcement, i.e. a reinforcement mechanism mainly based on the last observations, (ii) a random fluctuation of the conditional probabilities, and (iii) a long-term convergence of the empirical mean to a deterministic limit, together with a chi-squared goodness of fit result. This triple purpose has been achieved by the introduction of a new variant of the Eggenberger-Polya urn, that we call the "Rescaled" Polya urn. We provide a complete asymptotic characterization of this model and we underline that, for a certain choice of the parameters, it has properties different from the ones typically exhibited from the other urn models in the literature. As a byproduct, we also provide a Central Limit Theorem for a class of linear functionals of non-Harris Markov chains, where the asymptotic covariance matrix is explicitly given in linear form, and not in the usual form of a series.