Use mus17data_z.dta
* Summary stats for ER use model
global xlist1 age actlim totchr
summarize er $xlist1
tabulate er
* Here we are going to fit some zero-inflated models to the emergency room
* visits data
* Let us first fit the data with a Negative Binomial model to see
* if there appears to be overdispersion. Look at the estimate
* of the alpha coefficient and its significance
nbreg er age actlim totchr
* Since overdispersion is present in the data, we will pursue the
* Negative Binomial model
* Zero-Inflated Negative Binomial for er
zinb er $xlist1, inflate($xlist1) vuong nolog
* The Voung test statistic (which has a standard normal distribution)
* helps us to determine the preferability of the NB model versus the
* zero-inflated NB model. If the test statistic is significantly
* positive, the Zero-Inflated NB model is favored. On the other hand, if
* the test statistic is significantly negative, the NB model is to be
* preferred. If the test statistic is not significant negatively or
* positively, the test outcome is inconclusive. In the ER data the
* Voung statistic is signficantly positive with a one-tail p-vaule of 0.0233
* indicating a preference for the Zero-Inflated NB model.