Logistic regression on Markov chains for crop rotation modelling.

摘要

Often, in dynamical systems, such as farmer's crop choices, the dynamics is driven by external non-stationary factors, such as rainfall, temperature, and economy. Such dynamics can be modelled by a non-stationary Markov chain, where the transition probabilities are logistic functions of such external factors. We investigate the problem of estimating the parameters of the logistic model from data, using conjugate analysis with a fairly broad class of priors, to accommodate scarcity of data and lack of strong prior expert opinions. We show how maximum likelihood methods can be used to get bounds on the posterior mode of the parameters.