Detecting over- and under-dispersion in zero inflated data with the hyper-Poisson regression model