From s3460743 at student.rmit.edu.au  Tue Apr 11 11:46:11 2017
From: s3460743 at student.rmit.edu.au (Sivanandan Balakrishnan)
Date: Tue, 11 Apr 2017 11:46:11 +1000
Subject: [Limdep Nlogit List] Could Generalized ordered logit be analysed in
	Nlogit 5
Message-ID: <CAF5LGpOVf6_5mJ1Onf1d5zGR44=fM9KQK7nQj2xSxk45hP+roQ@mail.gmail.com>

Dear Nlogit users,

Could Generalized ordered logit be analysed or estimated in Nlogit 5 ?
If yes, how ?,

regards,
Sivanandan Balakrishnan
PhD Student in Civil Engineering *(Transport and Traffic Engineering)*,
School of Engineering,
RMIT University,
GPO BOX 2476,
Melbourne, Victoria 3001,
Australia.
Phone:Tel: +61 3 9925 3848
Email:s3460743 at student.rmit.edu.au

From christopher.standen at sydney.edu.au  Wed Apr 12 15:09:16 2017
From: christopher.standen at sydney.edu.au (Christopher Standen)
Date: Wed, 12 Apr 2017 05:09:16 +0000
Subject: [Limdep Nlogit List] NLOGIT ;simulate
Message-ID: <2382DFB83BB6E14F8E87F7EE08842B1101107432FF@ex-mbx-pro-05>

Hello,

I'm looking for some help trying to understand how the NLOGIT ;simulate subcommand works. I can't find any information in the documentation.

In particular:
-       How are the market shares calculated?
-       How many iterations are used for the simulation?
-       For random parameters, does it use conditional or unconditional estimates?
-       If unconditional, does it use the mean for every observation, or draw randomly from the distribution?
-       How are error components handled?


CHRIS STANDEN | PhD Candidate
Institute of Transport and Logistics Studies | The University of Sydney Business School

THE UNIVERSITY OF SYDNEY
Building H73, The University of Sydney NSW 2006
M +61 4 3177 6255 | F +61 2 9114 1683
E christopher.standen at sydney.edu.au<mailto:christopher.standen at sydney.edu.au>  | W http://sydney.edu.au/business/itls/staff/christopher.standen

Celebrating 25 years of ITLS: 1991-2016 http://youtu.be/s2D0T1crZwY
ERA Rank 5 (Transportation and Freight Services)
Join the ITLS group on LinkedIn<http://www.linkedin.com/groups?gid=682757>




From medakseth at gmail.com  Wed Apr 12 16:40:46 2017
From: medakseth at gmail.com (medard kakuru)
Date: Wed, 12 Apr 2017 09:40:46 +0300
Subject: [Limdep Nlogit List] Data set problem
Message-ID: <CALwnpeTUGGJhcCeZ1=U2ac1ApT+R2pfg4OrNiEUkHmMUma7AHg@mail.gmail.com>

Dear Users, I am facing a challenge with re-running an analysis I did 3
years a go. my data set has 540 observations. Three years a go, I run a MNL
model with no challenges. When I trying running the same model on the same
data set, the output shows that only 180 observations are used. I have
checked my data set over and over again but don?t see any problem. kindly
help me out.
Thanks.

From wgreene at stern.nyu.edu  Wed Apr 12 21:45:29 2017
From: wgreene at stern.nyu.edu (William Greene)
Date: Wed, 12 Apr 2017 07:45:29 -0400
Subject: [Limdep Nlogit List] Data set problem
In-Reply-To: <CALwnpeTUGGJhcCeZ1=U2ac1ApT+R2pfg4OrNiEUkHmMUma7AHg@mail.gmail.com>
References: <CALwnpeTUGGJhcCeZ1=U2ac1ApT+R2pfg4OrNiEUkHmMUma7AHg@mail.gmail.com>
Message-ID: <CAEJPMR6-q_8FM0mHoTMvJAVsmM7dDMyEoXkhUTxM697ib2Fg5A@mail.gmail.com>

If you are fitting an MNL with 3 alternatives and you have 540 rows of data,
then you have 180 observations and each one is a block of 3 rows of the
data.  That is what limdep is reporting.  If this is correct, there is
nothing
wrong with your data.  The 180 refers to blocks of 3 "observations."
/Bill Greene

On Wed, Apr 12, 2017 at 2:40 AM, medard kakuru <medakseth at gmail.com> wrote:

> Dear Users, I am facing a challenge with re-running an analysis I did 3
> years a go. my data set has 540 observations. Three years a go, I run a MNL
> model with no challenges. When I trying running the same model on the same
> data set, the output shows that only 180 observations are used. I have
> checked my data set over and over again but don?t see any problem. kindly
> help me out.
> Thanks.
> _______________________________________________
> Limdep site list
> Limdep at limdep.itls.usyd.edu.au
> http://limdep.itls.usyd.edu.au




-- 
William Greene
Department of Economics
Stern School of Business, New York University
44 West 4 St., 7-90
New York, NY, 10012
URL: http://people.stern.nyu.edu/wgreene
Email: wgreene at stern.nyu.edu
Ph. +1.212.998.0876

From wgreene at stern.nyu.edu  Wed Apr 12 23:42:53 2017
From: wgreene at stern.nyu.edu (William Greene)
Date: Wed, 12 Apr 2017 09:42:53 -0400
Subject: [Limdep Nlogit List] NLOGIT ;simulate
In-Reply-To: <2382DFB83BB6E14F8E87F7EE08842B1101107432FF@ex-mbx-pro-05>
References: <2382DFB83BB6E14F8E87F7EE08842B1101107432FF@ex-mbx-pro-05>
Message-ID: <CAEJPMR6pzDUBfAhYEzMK3qtxX_W=ExkF21i=qz4xkZaTOGyY5A@mail.gmail.com>

Chris:
1.  The market shares are computed by computing and summing the
probabilities.
2.  The computation is not iterative.
3/4.  The parameters in an RP model are computed as they would be during
the estimation
iterations.  The probabilities are simulated by drawing from the underlying
distribution.
5. Error components are handled the same as random parameters.
3/4/5.  For the simulation, the probabilities are computed as if the
likelihood were being
computed - the data setup for the simulation scenario is made immediately
before the
probabilities are computed - e.g., if you specify a price to rise by 10% as
the scenario, the
entire data setup is done with the original data, then immediately before
computing the
probabilities, the data for the observation are suitably modified.
/Bill Greene

On Wed, Apr 12, 2017 at 1:09 AM, Christopher Standen <
christopher.standen at sydney.edu.au> wrote:

> Hello,
>
> I'm looking for some help trying to understand how the NLOGIT ;simulate
> subcommand works. I can't find any information in the documentation.
>
> In particular:
> -       How are the market shares calculated?
> -       How many iterations are used for the simulation?
> -       For random parameters, does it use conditional or unconditional
> estimates?
> -       If unconditional, does it use the mean for every observation, or
> draw randomly from the distribution?
> -       How are error components handled?
>
>
> CHRIS STANDEN | PhD Candidate
> Institute of Transport and Logistics Studies | The University of Sydney
> Business School
>
> THE UNIVERSITY OF SYDNEY
> Building H73, The University of Sydney NSW 2006
> M +61 4 3177 6255 | F +61 2 9114 1683
> E christopher.standen at sydney.edu.au<mailto:christopher.
> standen at sydney.edu.au>  | W http://sydney.edu.au/business/
> itls/staff/christopher.standen
>
> Celebrating 25 years of ITLS: 1991-2016 http://youtu.be/s2D0T1crZwY
> ERA Rank 5 (Transportation and Freight Services)
> Join the ITLS group on LinkedIn<http://www.linkedin.com/groups?gid=682757>
>
>
>
> _______________________________________________
> Limdep site list
> Limdep at limdep.itls.usyd.edu.au
> http://limdep.itls.usyd.edu.au
>
>


-- 
William Greene
Department of Economics
Stern School of Business, New York University
44 West 4 St., 7-90
New York, NY, 10012
URL: http://people.stern.nyu.edu/wgreene
Email: wgreene at stern.nyu.edu
Ph. +1.212.998.0876

From carrolp7 at tcd.ie  Thu Apr 13 20:00:55 2017
From: carrolp7 at tcd.ie (Paraic Carroll)
Date: Thu, 13 Apr 2017 11:00:55 +0100
Subject: [Limdep Nlogit List] Status quo/ no-choice option SP data set-up
Message-ID: <CAFha20RvWtSq_3aq0wjO7KFqLkpQoUN1U3f_5BcBu1xz=ysGCg@mail.gmail.com>

Hi there,

I am conducting a SP mode choice experiment in which there are 3 modes/
alternatives (car, walk, cycle). Walk and cycle have three attributes with
3 levels each (infrastructure, time and adjacent traffic speed) that are
improved to various degrees by a range of policy measures in the scenarios.
However, car is being treated as a 'status quo' option that has no
attributes associated with it.

I have run a couple basic MNL models that have produced poor outputs in
terms of high p-values and coefficients that are not significant which
seems to suggest that maybe I have the dataset coded wrong.

I have been following the Shoe Brand Choice example conducted by William
Greene and I have tried to set the data up in the same way (e.g. fashion,
quality, price, asc4). However as my attributes are on three levels and not
two like the fashion and quality attributes I am unsure how to code them,
also in the shown dataset for this experiment price is coded 0.12, 0.08,
0.2 etc. and not 1,2,3,4 which confused me.

I have tried design coding (1,2,3), effects coding (-1,0,1) as well as
using the numerical values from the survey for the 3 attiribute levels in
each attribute, leaving the car option with a 0 constant and I have created
an ASC variable with 1 = car 'status quo' and 0 = other alternatives.

I am wondering if I am doing the right thing here or if I have made an
error in how I have prepared the data. I am also thinking that maybe a
nested model would be more suited to what I am trying to model.

Any suggestions would be greatly appreciated,

Kind regards,

P?raic

From yang.yang at usask.ca  Sat Apr 22 08:08:10 2017
From: yang.yang at usask.ca (Yang, Yang)
Date: Fri, 21 Apr 2017 22:08:10 +0000
Subject: [Limdep Nlogit List] RPLOGIT does not allow for proper
	heterogeneity in the variances
Message-ID: <1492812490430.34556@usask.ca>

?Dear Nlogit Users,


I estimated a random parameter logit model using Nlogit 6. In total, there are 6 random parameters, with 3 of them were specified as heterogeneous around the means and variances and the rest 3 were homogeneous and homoscedastic. However, the combination of binary values in fcn(# !) does not give proper estimation results. The codes and results are shown in the P.S.


In my case, it seems like "#" works well, but the "!" does not provide correct parameter estimates, i.e., parameters specified with fcn(!00) are still heteroscedastic. I appreciate if anyone can advise what may cause the problem and how to fix it. Thank you very much!


Yang


P.S.


SAMPLE  ; all $
TIMER $
CALC  ; ran(712) $
RPLOGIT  ; lhs = choice, cset, alti
                 ; choices = A, B, C, D
                 ; pds = 6
                 ; model:u(A) = 0 + pnb*nb + pae*ae + pge*pm1 + pgm*pm2 + pec*pm3 + ppri*pri /
                                u(B) = 0 + pnb*nb + pae*ae + pge*pm1 + pgm*pm2 + pec*pm3 + ppri*pri /
                                u(C) = 0 + pnb*nb + pae*ae + pge*pm1 + pgm*pm2 + pec*pm3 + ppri*pri /
                                u(D) = d
                 ; output=IC
                 ; halton
                 ; pts = 100
                 ; maxit = 200
                 ; rpl = b_no, bp_narr
                 ; hfr = b_no, bp_narr
                 ; fcn = pnb(n|#00!00), pae(n|#00!00), pge(n|#11!11), pgm(n|#11!11), pec(n|#11!11), ppri(o|#00!00) ?
                 ; par
                 ; wtp = pnb/ppri, pae/ppri, pge/ppri, pgm/ppri, pec/ppri  $




       |                  Standard            Prob.      95% Confidence
  CHOICE|  Coefficient       Error       z    |z|>Z*         Interval
--------+--------------------------------------------------------------------
        |Random parameters in utility functions..............................
     PNB|     .73332***      .19034     3.85  .0001      .36026   1.10638
     PAE|    -.02348         .16747     -.14  .8885     -.35171    .30474
     PGE|   -1.66754***      .25230    -6.61  .0000    -2.16204  -1.17304
     PGM|   -2.46845***      .26151    -9.44  .0000    -2.98099  -1.95590
     PEC|   -2.53523***      .25533    -9.93  .0000    -3.03567  -2.03479
    PPRI|   -1.89376***      .06111   -30.99  .0000    -2.01354  -1.77398
        |Nonrandom parameters in utility functions...........................
       D|   -9.59739***      .27756   -34.58  .0000   -10.14140  -9.05337
        |Heterogeneity in mean, Parameter:Variable...........................
 PNB:B_N|        0.0    .....(Fixed Parameter).....
 PNB:BP_|        0.0    .....(Fixed Parameter).....
 PAE:B_N|        0.0    .....(Fixed Parameter).....
 PAE:BP_|        0.0    .....(Fixed Parameter).....
 PGE:B_N|    -.68219**       .33986    -2.01  .0447    -1.34830   -.01608
 PGE:BP_|    1.00234***      .29651     3.38  .0007      .42120   1.58348
 PGM:B_N|    -.61873*        .35717    -1.73  .0832    -1.31877    .08131
 PGM:BP_|     .68697**       .28617     2.40  .0164      .12609   1.24786
 PEC:B_N|     .68527**       .29930     2.29  .0220      .09866   1.27188
 PEC:BP_|     .26818         .34257      .78  .4337     -.40325    .93961
PPRI:B_N|        0.0    .....(Fixed Parameter).....
PPRI:BP_|        0.0    .....(Fixed Parameter).....
        |Distns. of RPs. Std.Devs or limits of triangular....................
   NsPNB|    3.63726***      .25251    14.40  .0000     3.14235   4.13218
   NsPAE|    2.51460***      .20341    12.36  .0000     2.11591   2.91328
   NsPGE|    1.88191***      .23768     7.92  .0000     1.41606   2.34776
   NsPGM|    1.76437***      .23479     7.51  .0000     1.30418   2.22456
   NsPEC|    1.60885***      .22581     7.12  .0000     1.16627   2.05143
  TsPPRI|    1.89376***      .06111    30.99  .0000     1.77398   2.01354
        |Heteroscedasticity in random parameters.............................
 sPNB|B_|    -.23051**       .10233    -2.25  .0243     -.43108   -.02994
 sPNB|BP|    -.10699         .09936    -1.08  .2816     -.30173    .08776
 sPAE|B_|    -.03226         .13381     -.24  .8095     -.29451    .23000
 sPAE|BP|     .05959         .11095      .54  .5912     -.15786    .27704
 sPGE|B_|    -.03325         .19776     -.17  .8665     -.42086    .35435
 sPGE|BP|    -.29729         .22377    -1.33  .1840     -.73587    .14128
 sPGM|B_|    -.42754         .31790    -1.34  .1787    -1.05061    .19554
 sPGM|BP|   -29.1298      .6985D+12      .00 1.0000 ***********  ***********
 sPEC|B_|    -.44708         .33378    -1.34  .1804    -1.10129    .20712
 sPEC|BP|     .49542***      .17826     2.78  .0054      .14604    .84479
sPPRI|B_|     .34530***      .08448     4.09  .0000      .17971    .51088
sPPRI|BP|     .17615**       .07949     2.22  .0267      .02036    .33194
--------+--------------------------------------------------------------------



________________________________
Yang Yang, PhD candidate
Department of Agricultural and Resource Economics
College of Agriculture and Bioresources
University of Saskatchewan, Canada
Email: yang.yang at usask.ca







From fdzanku at gmail.com  Tue Apr 25 02:35:08 2017
From: fdzanku at gmail.com (Fred Dzanku)
Date: Mon, 24 Apr 2017 16:35:08 -0000
Subject: [Limdep Nlogit List] Marginal Effects after Panel Data Bivariate
	Probit
Message-ID: <011601d2bd18$be62daa0$3b288fe0$@gmail.com>

Does Nlogit 4 provide marginal effects after estimating RE bivariate probit?
Fred 

-----Original Message-----
From: limdep-bounces at limdep.itls.usyd.edu.au
[mailto:limdep-bounces at limdep.itls.usyd.edu.au] On Behalf Of William Greene
Sent: Wednesday, April 12, 2017 1:43 PM
To: Limdep and Nlogit Mailing List <limdep at limdep.itls.usyd.edu.au>
Subject: Re: [Limdep Nlogit List] NLOGIT ;simulate

Chris:
1.  The market shares are computed by computing and summing the
probabilities.
2.  The computation is not iterative.
3/4.  The parameters in an RP model are computed as they would be during the
estimation iterations.  The probabilities are simulated by drawing from the
underlying distribution.
5. Error components are handled the same as random parameters.
3/4/5.  For the simulation, the probabilities are computed as if the
likelihood were being computed - the data setup for the simulation scenario
is made immediately before the probabilities are computed - e.g., if you
specify a price to rise by 10% as the scenario, the entire data setup is
done with the original data, then immediately before computing the
probabilities, the data for the observation are suitably modified.
/Bill Greene

On Wed, Apr 12, 2017 at 1:09 AM, Christopher Standen <
christopher.standen at sydney.edu.au> wrote:

> Hello,
>
> I'm looking for some help trying to understand how the NLOGIT 
> ;simulate subcommand works. I can't find any information in the
documentation.
>
> In particular:
> -       How are the market shares calculated?
> -       How many iterations are used for the simulation?
> -       For random parameters, does it use conditional or unconditional
> estimates?
> -       If unconditional, does it use the mean for every observation, or
> draw randomly from the distribution?
> -       How are error components handled?
>
>
> CHRIS STANDEN | PhD Candidate
> Institute of Transport and Logistics Studies | The University of 
> Sydney Business School
>
> THE UNIVERSITY OF SYDNEY
> Building H73, The University of Sydney NSW 2006 M +61 4 3177 6255 | F 
> +61 2 9114 1683 E 
> christopher.standen at sydney.edu.au<mailto:christopher.
> standen at sydney.edu.au>  | W http://sydney.edu.au/business/ 
> itls/staff/christopher.standen
>
> Celebrating 25 years of ITLS: 1991-2016 http://youtu.be/s2D0T1crZwY 
> ERA Rank 5 (Transportation and Freight Services) Join the ITLS group 
> on LinkedIn<http://www.linkedin.com/groups?gid=682757>
>
>
>
> _______________________________________________
> Limdep site list
> Limdep at limdep.itls.usyd.edu.au
> http://limdep.itls.usyd.edu.au
>
>


--
William Greene
Department of Economics
Stern School of Business, New York University
44 West 4 St., 7-90
New York, NY, 10012
URL: http://people.stern.nyu.edu/wgreene
Email: wgreene at stern.nyu.edu
Ph. +1.212.998.0876
_______________________________________________
Limdep site list
Limdep at limdep.itls.usyd.edu.au
http://limdep.itls.usyd.edu.au