[Limdep Nlogit List] Re: Another selection qeustion-with censoredmodel
Martin Chun Qiu
cqiu at ualberta.ca
Tue Oct 23 23:07:51 EST 2007
Thanks for those who responded.
Let me describe the sequence of the action. a decision maker has some factors v to consider the following DM problem.
Based on a subset of v, she decides to provide x or not (characterized by z). If she does (z=1), she sets x (exogenously given), and then she decides the level of y (as a function of v, ) which should not exceed x (y=max(bv, x)).
If she does not provide x (z=0), then she still needs to decide the level of y, which is unbounded (y=bv). In all cases, z and v are observed, x is only observed when z=1, and y is observed but with some censored quantities.
How to model this problem?
Warm regards,
Chun (Martin) Qiu
======= 2007-10-23 01:04:27 Your message=======
From: Fred Feinberg
Subject: Re: [Limdep Nlogit List] Another selection qeustion-with censoredmodel
>Can you clarify? Forgetting about Z totally for the moment, it sounds like, in the second equation, you never observe both X and Y together, for any observations. That is, when X is "absent", you observe Y, but when X is "present", you don't observe Y. I'm not even sure how one would write a likelihood for that one
>equation, even if its error term was uncorrelated with the model for Z. Was there a typo here?
>
>FF
>
>Martin Chun Qiu wrote:
>
>> Dear Limdepers:
>>
>> I am also estimating a selection problem where the first equation is a standard probit on z. In the second equation, the dependent variable y is censored by x. The presence of x is determined by z. If z=1, x is present and y is censored. If z=0, x is absent and y is not censored. Any thoughts to handle model like this?
>>
>> Warm regards,
>>
>> Chun (Martin) Qiu
>> Desautels Faculty of Management
>> Montreal, QC
>> Canada H3A 1G5
>>
>> _______________________________________________
>> Limdep site list
>> Limdep at limdep.itls.usyd.edu.au
>> http://limdep.itls.usyd.edu.au
>
>
>
>_______________________________________________
>Limdep site list
>Limdep at limdep.itls.usyd.edu.au
>http://limdep.itls.usyd.edu.au
>
= = = = = = = = = = = = = = = = = = = =
Can you clarify? Forgetting about Z totally for the moment, it sounds like, in the second equation, you never observe both X and Y together, for any observations. That is, when X is "absent", you observe Y, but when X is "present", you don't observe Y. I'm not even sure how one would write a likelihood for that one
equation, even if its error term was uncorrelated with the model for Z. Was there a typo here?
FF
Martin Chun Qiu wrote:
> Dear Limdepers:
>
> I am also estimating a selection problem where the first equation is a standard probit on z. In the second equation, the dependent variable y is censored by x. The presence of x is determined by z. If z=1, x is present and y is censored. If z=0, x is absent and y is not censored. Any thoughts to handle model like this?
>
> Warm regards,
>
> Chun (Martin) Qiu
> Desautels Faculty of Management
> Montreal, QC
> Canada H3A 1G5
>
> _______________________________________________
> Limdep site list
> Limdep at limdep.itls.usyd.edu.au
> http://limdep.itls.usyd.edu.au
_______________________________________________
Limdep site list
Limdep at limdep.itls.usyd.edu.au
http://limdep.itls.usyd.edu.au
More information about the Limdep
mailing list