Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down at the end of May, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: st: nested logit tree


From   "Alvarez,Sergio" <sergioal@ufl.edu>
To   <statalist@hsphsun2.harvard.edu>
Subject   RE: st: nested logit tree
Date   Mon, 28 Nov 2011 11:52:14 -0500

Makes sense.  Check this website:
http://www.stata.com/support/faqs/stat/nlogit.html
It's an example using the restaurant.dta file.

I think you want to have the R&D vs. no R&D choice in the 'type' category in the example. Then you have 4 choices in the lower branch, but only two for each nest:

                             All firms
             R&D               vs.           no R&D
buy with R&D | sell with R&D        buy no R&D | sell no R&D

And then you have a dummy variable that indicates the actual choice.

The trick is that you need the 4 observations for each firm, so if you don't yet have them you need to find a way to create them. I did it using the -reshape- command, but it depends on how your dataset is structured.

Sorry if I'm being unclear, it just challenging to explain this.

Sergio

On Mon, 28 Nov 2011 17:33:59 +0100, Solon Moreira wrote:
Hi Sergio,

Thanks for your email. In fact I want a decision model that has 4
possible outcomes: a buyer with R&D, a buyer without R&D, a supplier
with R&D and a supplier without R&D. So in the end I have 4 possible
alternatives in my data. I had a look in the restaurant data, but it
is not exactly what I want, I think I need 2 levels. I could also
calculate different logit models but I am not sure that is the best
solution. I am expecting that the effect of some of my covariates will
have a different direction in affecting the likelyhood of a firm
having R&D or not depending on the side of the market that it it
(buyer of supplier).

Thanks,
Solon

________________________________________
From: owner-statalist@hsphsun2.harvard.edu
[owner-statalist@hsphsun2.harvard.edu] On Behalf Of Alvarez,Sergio
[sergioal@ufl.edu]
Sent: Monday, November 28, 2011 5:02 PM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: nested logit tree

Solon,

I may be misunderstanding, but it looks like you want to model a
decision which has only two possible outcomes: firm is a supplier or
firm is a buyer. If that is the case, you probably should use a probit
or logit rather than a nested logit.
You can also model whether the firm has an R&D department (yes or no)
using a similar approach.

The nested and conditional logits are used when you have an agent (a
firm) making a choice among many alternatives (more than two).

Take a look at the restaurant.dta sample dataset and the nlogit help
file and it will become clear.

Best,

Sergio

On Mon, 28 Nov 2011 14:39:19 +0100, Solon Moreira wrote:
Dear Stata list readers,

This is a question regarding the way that I should structure the tree
of a nested logit model to get the right output.

I have a dataset consisting of 1200 firms that buy and sell
technologies. I am trying to run a model in which in the first level
I
can observe how firm-specific variables affect the likelihood of
being
a supplier instead of a buyer. The way that my model is structured
now
this part of the nest is shown in my stata output with no problem.
However, I am having problem with my second level, I would like the
output to show how some alternative specific variables affect the
likelihood that buyers and suppliers will have an R&D department. So,
I want to produce an output in which the first level explains the
likelihood of being a buyer or a supplier and second level the
likelihood that the buyer will or not have an R&D department and the
supplier will or not have an R&D department. As I mentioned, I am
able
to produce an output in which I can observe the first level, but in
the second level instead of comparing buyers that have R&D with
buyers
that do not have it (and the same wit!
 h suppliers), my output shows all the 4 final alternatives in the
same model, taking one of them as the baseline.

That is the command I am using to build the tree:

nlogitgen r&d= final_outcome(buyer_n: 1, buyer_w: 2, supplier_n: 3,
supplier_w: 4)
nlogitgen buyer_or_supplier= r&d(buyer: buyer_n | buyer_w , supplier:
supplier_n | supplier_w)

I could not figure out a way to compare only buyers with and without
R&D and suppliers with and without R&D in the second level.

Any suggestions are more than welcome
Very best,
Solon

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

--
Sergio Alvarez
Food and Resource Economics
University of Florida
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

--
Sergio Alvarez
Food and Resource Economics
University of Florida
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index