Comments:
![]()
|
|
Date: 2019-03-20 07:25
Sender: Marek ChudyI have send you couple of emails, waiting for your reply. Thanks | Date: 2019-03-18 12:32
Sender: Arne HenningsenHi Marek
Information about SVN is available, e.g., at: http://svnbook.red-bean.com/
In order to access a SVN repository, you (only) need a SVN client (not the SVN server), e.g., the free 'foundation' edition of SmartSVN: https://www.smartsvn.com/
You can send me the file tobit2tobit1fit(.R?) by e-mail (you can find my e-mail address at: http://arne-henningsen.name).
Cheers,
Arne
| Date: 2019-03-17 15:37
Sender: Marek ChudyHi Arne,
I will request the membership in sampleSelection project as you suggest. I have no prior experience with SVN. As you say, I think the selection function would be the best choice for interface. The tobit2tobit1fit, which I will send you has the same structure as your tobit2Intfit procedure. I tried to upload the code in Attachments as text file yesterday, but it did not accept it. I do not have it with me now, but I will upload it tomorow morning, so that you can adjust it as neccessary. How sould I send it to you?
| Date: 2019-03-17 15:30
Sender: Arne HenningsenHi Marek!
Great that you implemented it! :-)
We use the "Subversion" (SVN) version control system to develop the source code of the "sampleSelection" package. Do you know "Subversion"?
Please go the R-Forge "summary" page of the "sampleSelection" package and click at "Request to join" in the "Project Members" box on the upper right so that you get write access to the source code of the sampleSelection package.
What is the 'user interface' of your implementation of a Tobit model with sample-selection? Perhaps, selection() should have two additional arguments, e.g., called "left" and "right" with default values -Inf and Inf, respectively. If the user sets argument "left" or argument "right" (or both arguments) to a finite number, selection() estimates a Tobit model with sample selection. What do you think? Do you have a better idea for a "user interface"?
Perhaps, you can send me your code -- preferably with a reproducible example -- so that I can investigate how to best integrate it into the "sampleSelection" package.
/Arne
| Date: 2019-03-16 15:29
Sender: Marek ChudyHi Arne,
I made the impementation based on your tobit2Intfit.R procedure. Please give me instructions how to add it to your package. I never did it, so any advice form you is most welcome.
Best
Marek | Date: 2019-03-15 07:50
Sender: Arne HenningsenHi Marek
You are right, the sampleSelection package cannot estimate sample-selection models with a Tobit / censored regression as outcome equation. sampleSelection can estimate sample-selection models with a binary dependent variable of the outcome equation (you could transform the dependent variable of the outcome equation to a binary variable but you would probably loose to much "information" when you do this) or an interval-coded dependent variable of the outcome equation (you could transform the dependent variable of the outcome equation to an interval-coded variable, where the lowest interval is -Inf to T, but depending on the distribution of the values of the dependent variable of your outcome equation, you could loose quite some information if you interval-code the dependent variable of your outcome equation). You are invited to implement the estimation of sample-selection regression with a Tobit / censored regression as outcome equation in the sampleSelection package. I could assist you with this.
You could also check whether the "mhurdle" package [1] can estimate a model specification that is suitable for your analysis
https://cran.r-project.org/package=mhurdle
Best regards,
Arne
|
|
