I have panel data with a skewed non-negative dependent variable, with a large mass at zero. I have used xsmle successfully to estimate SAR and SEM models, but so far have ignored the skewed nature of the data. I have also ignored the fact that the dependent variable is a count (integer-valued variable).
I would like to consider models that specifically accommodate these properties of the dependent variable. In this regard I would like to consider Negative Binomial (NBD) or Zero-inflated Poisson (ZIP) models, or even a Tobit model with a lower-threshold at zero (I recognize this model would treat the dependent variable as continuous instead of integer-valued). However, I do not believe xsmle can estimate these models.
I would be grateful for suggestions on how to estimate models like the NBD, ZIP, or Tobit on panel data with spatial dependence structures. I would also appreciate suggestions on other ways to accomplish what I am trying to do within STATA.
Thank you
I would like to consider models that specifically accommodate these properties of the dependent variable. In this regard I would like to consider Negative Binomial (NBD) or Zero-inflated Poisson (ZIP) models, or even a Tobit model with a lower-threshold at zero (I recognize this model would treat the dependent variable as continuous instead of integer-valued). However, I do not believe xsmle can estimate these models.
I would be grateful for suggestions on how to estimate models like the NBD, ZIP, or Tobit on panel data with spatial dependence structures. I would also appreciate suggestions on other ways to accomplish what I am trying to do within STATA.
Thank you
