I am trying to understand if I should use truncated negative binomial for my research. 50 percent of my dependent variable is zero. I already used negative binomial but I would like to use another model as well. So, I was wondering if someone could tell me if truncated negative binomial is a proper model for my dataset (my dataset includes 162 observations).
Asked
Active
Viewed 95 times
2
-
Unless you provide some additional information, this question is unanswerable: the choice of a model is not solely determined by the proportion of zeros in the dependent variable and the number of observations (even though it's true this information will slightly restrict the options). It depends on how that variable is related to the explanatory variables, as well as on what your data mean and your objectives for this analysis. – whuber Apr 19 '18 at 16:25
-
1Thank you for the reply. I want to analyze the number of people who joined violent groups per county. My independent variables include literacy rate, population, urbanization, unemployment rate, and whether the county includes ethnic minorities. DV is the number of individuals per county who joined those groups. – Susan mohammad.k Apr 19 '18 at 16:33
-
4I am uncertain what is meant by the term "truncated" in this context. This seems to be a zero-inflated process (where there is some factor determining if there is a chance of joining for the entire county...if not, zero; if yes, then the count follows the neg-binom). – Gregg H Apr 19 '18 at 16:43
-
It's unclear exactly what you mean by "analyze," but wouldn't you want, at the very minimum, to account for some measure of county population in your analysis? – whuber Apr 19 '18 at 17:24