[Limdep Nlogit List] Varying choice set by individual and time (yes/no question)

William Greene wgreene at stern.nyu.edu
Mon Dec 1 14:26:43 EST 2008

In the NLOGIT setup for panel data style estimators such as the
random parameters logit model, you could accommodate the scenario
you describe by using the appropriate ;PDS=number of choice sets.
The variable specified by the PDS command takes the same value for
every choice and for every repetition for the person.  For example,
if you had two individuals each choosing from the same 4 alt choice
set, the first one 3 times, the second 4 times, then the PDS variable
would equal 3 for the first 12 rows of data and would equal 4 for the
next 16.
/B. Greene

----- Original Message -----
From: "hielke buddelmeyer" <hielke_buddelmeyer at hotmail.com>
To: limdep at limdep.itls.usyd.edu.au
Sent: Sunday, November 30, 2008 7:58:59 PM GMT -05:00 Columbia
Subject: [Limdep Nlogit List] Varying choice set by individual and time (yes/no question)

Dear list,
I am 99 percent sure the answer is no, but I can't find explicit confirmation in the manual: Is there a panel version for the variable choice set specification that would let the choice set for individuals vary over time in a repeated choice setting? (A simple yes/no will do)
I know that the choice set can vary by individual (i.e. Lhs = y,nij,altij as outlined in chapter N10, bottom page N10-4), but what if the choice set varies over time, but not over individuals? That is, all individuals always face the same choice set in any given period, but the choice set varies over time. 
An example of a situation where this may appear is if transport options get expanded by the completion of a new piece of infrastructure. So in a panel of repeated choices, individuals are observed to take the bus, car, or bicycle but from a particular date onwards the same individuals are now also observed to use a new train service that came online. Another example could be labour market outcomes in Spain where from a particular date onwards fixed-term contracts became available. So pre-reform outcomes may have been unemployed, full-time permanent and part-time permanent. Post-reform additional outcomes are full-time on a fixed-term contract, or part-time on a fixed-term contract. And one has repeated observations for individuals spanning this pre and post-reform period. These are just examples. Of course the alternative would be to cut the sample in two to define a 'pre-reform' sample and a 'post-reform' sample and treat them independently. But I was just curious to see if 
 the choice set could be restricted over time instead.

Access your email online and on the go with Windows Live Hotmail.
Limdep site list
Limdep at limdep.itls.usyd.edu.au

More information about the Limdep mailing list