[Limdep Nlogit List] Nlogit simulation method

[William Greene]

Thao: The answer to your question depends on how you specified the
simulation. NLOGIT
carries out the simulation you specify it.  In general, the ;SIMULATE
option for the NLOGIT
or CLOGIT routine works as follows.
(1) The base case is not the ASC only model. (You could specify it to be -
I'll look at that below.)
The base case is the data as they are in the sample.  Probabilities are
computed for each
observation in the current sample using the model that has just been
estimated and the data
in the current sample.
(2) The simulation scenario consists of a set of one or more changes in
attributes. Step
(1) is repeated, but for each observation, the change in the attribute is
applied before the
probabilities are computed.  The change in the attribute may be a change,
such as a 1% change
in travel time, or it might be to fix an attribute to a particular value
for everyone.
The four part setup you list below is actually how NLOGIT does the
calculations, save for the
specification of the base case.  You can specify the fixed base case the
way you want as
follows: Note first that the simulator is worked on the data, any subset of
the data, or a
completely different set of observations.  So, to fix the base, you would
create a single
artificial observation that contains the precise base case you want.  You
can then run the
simulation using that single observations with a simulated scenario.  This
is rather like
computing partial effects, however, which can be done much easier using the
specific tools
for that purpose.  The simulator tries to predict the number of
observations in base and
changed cases, which would not be useful if you were just computing the
probabilities for
a single base case and a single change to that base.
Regards,
Bill Greene

On Mon, Nov 30, 2020 at 5:25 AM Thao Thai via Limdep <
limdep at mailman.sydney.edu.au> wrote:

> Hi Nlogit users,
>
> I would be grateful if you could  explain how Nlogit undertakes the
> simulation method which is descried in nlogit manual:
>  (1) base case is ASC only model (which means all attributes take on Zero)
>  (2) Change one or some  attribute value
>  (3) Here is what I am confused about: *How did Nlogit arrives the
> probabilities of choosing for each alternative *
>      - Assume other attributes takes zero except changed attribute take the
> values specified in (2)
> OR
>     - Other attributes take values as is in the data and the changed
> attribute take the new value specified in (2)
>
> This is because I am not sure if the simulation described in Nlogit is
> similar to the "predictive uptake analysis" usually done in Health
> Economics literature and they call it "simulation". The process is:
>          (1) specify a base case in which all attributes take on fixed
> values and calculate probabilities of choosing each alternative
>           (2) change values of one or some attributes
>           (3) recalculate the probabilities of choosing each alternative
>           (4) the differences between probabilities of choosing the same
> alternative before and after are attributable to the change of
> attribute values.
>
>
> My second question: *How does Nlogit undertake the simulation using a mixed
> logit model?* Does it simply use the mean value of coefficients or does
> nlogit incorporate the distribution of coefficients in their calculation?
>
> Thank you so much. I look forward to seeing your explanation.
> Thao
> --
> *Thao Thai*| MHEcon(Adv), BPharm
> PhD candidate
> Centre for Health Economics <https://protect-au.mimecast.com/s/rmWTC6XQ4LfgD6YYfpIKPx?domain=monash.edu>
> Monash Business School | Monash University
> Tel (03) 99029847| Thao.T.Thai at monash.edu  <Thao.T.Thai at monash.edu>
> _______________________________________________
> Limdep site list
> Limdep at mailman.sydney.edu.au
> https://protect-au.mimecast.com/s/J_3SC71R2NT0qX11sWDOIV?domain=limdep.itls.usyd.edu.au
>
>

-- 
William Greene
Department of Economics, emeritus
Stern School of Business, New York University
44 West 4 St.
New York, NY, 10012
URL: https://protect-au.mimecast.com/s/HGRKC81V0PTAr5ppS2eyw_?domain=people.stern.nyu.edu
Email: wgreene at stern.nyu.edu
Editor in Chief: Journal of Productivity Analysis
<https://protect-au.mimecast.com/s/YApxC91WPRT61WEEiOC0h4?domain=springer.com>
<https://protect-au.mimecast.com/s/YApxC91WPRT61WEEiOC0h4?domain=springer.com>
Editor in Chief: Foundations and Trends in Econometrics
Associate Editor: Economics Letters
Associate Editor: Journal of Business and Economic Statistics