Forum: open-discussion

RE: Run dummy variable z determinant efficiency with Frontier 4.1 [ Reply ] By: Dhanu Utomo on 2016-06-04 16:43	[forum:43258] aso.txt (9) downloads
Dear Prof Arne, I want to ask, why my all variable in mle result have standard error 1. Thank you Prof Arne for your kindness reply.

RE: Run dummy variable z determinant efficiency with Frontier 4.1 [ Reply ]
By: Dhanu Utomo on 2016-05-01 11:38

[forum:43173]

My result of likelihood ratio with Cobb-Douglas function is 311.3. Which is larger than critical value with degree of value 15 in chi square = 24.996, 15 is because the different variable use in model translog and Cobb-Douglas is 15. Which is 20 - 5 = 15.

For my problem, with the ML estimation is less than that obtained using OLS, i read your comment in other thread, which say,"In order to rule this out, you could restart the estimation with user-defined starting values (perhaps slightly different from the automatically chosen starting values) and to decrease the tolerance level for detecting convergence."

What is the starting value, and how i can use it to run in Front 4.1.

Thank you Prof Arne for your reply and kindness.

RE: Run dummy variable z determinant efficiency with Frontier 4.1 [ Reply ]
By: Arne Henningsen on 2016-04-30 20:45

[forum:43172]

Do I correctly understand that the likelihood ratio test indicates that the fit of a Translog SFA model is not significantly better than the fit of a Translog OLS model?

Theoretically, the likelihood value of the ML estimation of an SFA model cannot be smaller than the likelihood value of the same model estimated by OLS. If you observe this, it seems that the ML estimation procedure did not find the maximum of the likelihood value. Please check this.

Do I also correctly understand that the likelihood ratio test indicates that the fit of a Translog SFA model is significantly better than the fit of a Cobb-Douglas SFA model?

In this case, you should *not* use the Cobb-Douglas model.

RE: Run dummy variable z determinant efficiency with Frontier 4.1 [ Reply ]
By: Dhanu Utomo on 2016-04-29 12:15

[forum:43170]

Thank your Prof Arne for your reply. That's really help me. Thank you, thank you very much.

I have other question, when i run one of my model with Cobb-Douglas function. The likelihood ratio suggest if there are technical ineffciency and suggest if the translog funtion is the best model to be use. But when i run it again with translog function, there are no technical efficiency in model, because the likelihood value is less than that obtained using ols. What is wrong with that? Or can I just use the Cobb-Douglas function other than translog function?

RE: Run dummy variable z determinant efficiency with Frontier 4.1 [ Reply ]
By: Arne Henningsen on 2016-04-20 10:37

[forum:43155]

Dear Dhanu

It seems to me that your 2 variables regarding ownership type and crisis are *not* dummy variables but *categorical* variables (called "factors" in R) with each of them having 3 distinct values (called "levels" in R). Hence, you need 2 dummy variable (0-1 variables) for each categorical variable, i.e. in total 4 dummy variables. If your two categorical variables are defined as "factors" in R, sfa() automatically creates the 4 required dummy variables. If your data set ("data.frame" in R) has the name, say, "myDat" and the categorical variables have the names, say, "owner" and "crisis", the commands "class( myDat$owner )" and "class( myDat$crisis )" both should return "factor". Otherwise, you have to convert these variables to factors, e.g. using the command "factor" or "as.factor". Then you can estimate an SFA model using the command:

sfa( y ~ x1 + x2 + owner + crisis, data = myDat )

or

sfa( y ~ x1 + x2 | owner + crisis, data = myDat )

where the first command estimates a model, where "owner" and "crisis" may affect the maximum achievable output quantity, while the second command estimates a model, where "owner" and "crisis" do *not* affect the maximum achievable output quantity, but may affect the distance between the maximum attainable output quantity and the observed output quantity.

Best regards,
Arne

Run dummy variable z determinant efficiency with Frontier 4.1 [ Reply ]
By: Dhanu Utomo on 2016-04-20 05:41

[forum:43151]

Data.xlsx (9) downloads

Dear Prof. Arne,

Can I regress z determinant dummy variable with frontier 4.1? I have 2 dummy variable. The first one is for ownership type, and the other is for crisis. In ownership type, i have 3 category, state own, commercial own, and foreign own. And the second dummy is for crisis 2005, crisis 2008, and non crisis. I already try to run it by value 1, if commercial own, zero otherwise. value 1 if foreign own, zero otherwise. And state own is the benchmark. And 1 for crisis 2005, zero otherwise and also in crisis 2008. Non-crisis as the benchmark. Is it ok to do that?

I really need your help. Please enlighten me, Prof Arne.