Hello everyone.
I am working with a count-dependent variable that takes on values 0,1,2,.......,11,12. I have panel data for 3 years. I am unable to decide whether the Poisson model or negative binomial is a good fit. I saw somewhere to use the negative binomial if the standard deviation is greater than 1 or in some places if the mean and variances are not equal. The mean and standard deviation of my dependent variable y are 9.987459 and 1.548862 respectively. I have tried to fit xtpoisson with fixed effect and it works. But with xtnbreg my model doesn't converge at all, it carries out iteration after iteration.
I don't know which model should I use. Any suggestion is appreciated.
I am working with a count-dependent variable that takes on values 0,1,2,.......,11,12. I have panel data for 3 years. I am unable to decide whether the Poisson model or negative binomial is a good fit. I saw somewhere to use the negative binomial if the standard deviation is greater than 1 or in some places if the mean and variances are not equal. The mean and standard deviation of my dependent variable y are 9.987459 and 1.548862 respectively. I have tried to fit xtpoisson with fixed effect and it works. But with xtnbreg my model doesn't converge at all, it carries out iteration after iteration.
I don't know which model should I use. Any suggestion is appreciated.
Comment