Hi, I'm new here.
I'm currently using churdle command to implement double hurdle model, and I found out that the log likelihood function built in this command is the one that first decides whether to adopt and then the amount(or extent) of it, by reading Cragg(1971) -Some statistical models for limited dependent variables with application to the demand for durable goods-.
In his paper, he suggests two types of decision processes: Firstly, desire positive acquisition and then carry out the adjustment. Secondly, a decision first has to be made about whether to consider a change or not, and then a decision on the amount of the change is taken. In the end, he proves that the both models can be represented by the same likelihood function by adding another additional assumption in the later model.
Anyway, the likelihood function introduced in churdle description is the later one.
and I'm facing three problems:
1. In another paper( Jones, A. (1989). A Double-Hurdle Model of Cigarette Consumption. Journal of Applied Econometrics, 4(1), 23-39. Retrieved from http://www.jstor.org/stable/2096488), it says that "unlike the Tobit model or the dominance model introduced below, L does not separate into a standard probit and a truncated regression". However, the command description of churdle is saying "Separate independent covariates are permitted for each model" and it is! I tried probit and truncated regression separately with my data, and the estimators are truly the same! So my question is, is the command wrong in that it allows the separation of two models?
2. One distinct feature of double hurdle model is that it has two types of zero(or the truncated values): zeros at adoption and zeros at unbounded outcome. However, when I use churdle, it has one source of zero: zeros at adoption, because the second model is truncated regression. Even though I followed the proof from Cragg(1971) that the two types of decision processes are fundamentally equivalent, but I can't interpret in the same way! Is there anything I miss?
3. churdle command itself is Cragg's double hurdle model, but according to the stata paper - Engel and Moffat. (2014). dhreg, xtdhreg, and bootdhreg: Commands to implement double-hurdle regression.The Stata Journal 14, Number 4, pp. 778–797- , churdle should be the single-hurdle model because it has only one source of zeros. but I know that Cragg's model is not a first-hurdle dominance model. So my question is.....is churdle truly a double hurdle model? this question is still related to the second one, but I seriously need confirmation.
Thank you so much!!
I'm currently using churdle command to implement double hurdle model, and I found out that the log likelihood function built in this command is the one that first decides whether to adopt and then the amount(or extent) of it, by reading Cragg(1971) -Some statistical models for limited dependent variables with application to the demand for durable goods-.
In his paper, he suggests two types of decision processes: Firstly, desire positive acquisition and then carry out the adjustment. Secondly, a decision first has to be made about whether to consider a change or not, and then a decision on the amount of the change is taken. In the end, he proves that the both models can be represented by the same likelihood function by adding another additional assumption in the later model.
Anyway, the likelihood function introduced in churdle description is the later one.
and I'm facing three problems:
1. In another paper( Jones, A. (1989). A Double-Hurdle Model of Cigarette Consumption. Journal of Applied Econometrics, 4(1), 23-39. Retrieved from http://www.jstor.org/stable/2096488), it says that "unlike the Tobit model or the dominance model introduced below, L does not separate into a standard probit and a truncated regression". However, the command description of churdle is saying "Separate independent covariates are permitted for each model" and it is! I tried probit and truncated regression separately with my data, and the estimators are truly the same! So my question is, is the command wrong in that it allows the separation of two models?
2. One distinct feature of double hurdle model is that it has two types of zero(or the truncated values): zeros at adoption and zeros at unbounded outcome. However, when I use churdle, it has one source of zero: zeros at adoption, because the second model is truncated regression. Even though I followed the proof from Cragg(1971) that the two types of decision processes are fundamentally equivalent, but I can't interpret in the same way! Is there anything I miss?
3. churdle command itself is Cragg's double hurdle model, but according to the stata paper - Engel and Moffat. (2014). dhreg, xtdhreg, and bootdhreg: Commands to implement double-hurdle regression.The Stata Journal 14, Number 4, pp. 778–797- , churdle should be the single-hurdle model because it has only one source of zeros. but I know that Cragg's model is not a first-hurdle dominance model. So my question is.....is churdle truly a double hurdle model? this question is still related to the second one, but I seriously need confirmation.
Thank you so much!!
