I examine the deterrent effect of executions on homicides between 1979 and 1998 in the United States, by using the panel data. Because the state-day is my unit of analysis, my data set is simply large with 352,555 cases. And, the total number of variables is around 195. For your further information, I use the negative binomial regression model (xtnbreg) because the dependent variable is a count data and the conditional mean is greater than the conditional variance.
What makes me confused about the results of the Stata analysis is this: While my fixed-effect model can converge, my random-effect model cannot converge. I did not change the equation model between the fixed-effect model and the random-effect model. So, I used the maxiter command in order to estimate the random-effect model. And, the Hausman test shows that the random-effect model is more appropriate than the fixed-effect model.
Anyone should be appreciated if you can let me know why this happens and whether both fixed-effect and random-effect models can converge at the same time.
What makes me confused about the results of the Stata analysis is this: While my fixed-effect model can converge, my random-effect model cannot converge. I did not change the equation model between the fixed-effect model and the random-effect model. So, I used the maxiter command in order to estimate the random-effect model. And, the Hausman test shows that the random-effect model is more appropriate than the fixed-effect model.
Anyone should be appreciated if you can let me know why this happens and whether both fixed-effect and random-effect models can converge at the same time.
Comment