From jgiller at ucdavis.edu  Wed Feb 16 08:20:28 2022
From: jgiller at ucdavis.edu (James Giller)
Date: Tue, 15 Feb 2022 13:20:28 -0800
Subject: [Limdep Nlogit List] Is a multinomial logit model with selection
	applicable here?
Message-ID: <CAP_EMv38kXrqRUNQFcK06hRpxTvVPgcMkKaKPJAyArjA=RzT-Q@mail.gmail.com>

Hi,

My dependent variable is based on travel mode choice survey data, and it
has a value for "make no trip at all" as well as values for various modes,
car, transit, taxi, walking, etc. The mode is only observed if the survey
participant would make a trip, which made me consider a sample selection
model.

My dependent variable Y has the following values:
0 = transit
1 = car
2 = carpool
3 = taxi
4 = walking
5 = other
6 = no trip

I created another variable DOTRIP, which is 0 if Y = 6, and 1 otherwise.

I then tried to estimate the following model
PROBIT ; Lhs = DOTRIP ; Rhs = <explanators> ; Hold $
MLOGIT ; Lhs = Y ; Rhs = <explanators> ; Selection ; Pts = 25; Halton $

The above model on the data set as explained returned an error message that
there is insufficient variation in the dependent variable during the MLOGIT
part. I think this is because the value 6 (no trip) is possible for Y, but
none of the observations would be selected in the selection step.

Am I correct in specifying a model with sample selection in this case? Or
should I take a different approach to account for the individuals first
deciding whether or not to travel, then by which mode?

Is there some way to rearrange my data so that the above model will work?

Thank you.
James.

From wgreene at stern.nyu.edu  Thu Feb 17 01:51:34 2022
From: wgreene at stern.nyu.edu (William Greene)
Date: Wed, 16 Feb 2022 09:51:34 -0500
Subject: [Limdep Nlogit List] Is a multinomial logit model with
 selection applicable here?
In-Reply-To: <CAP_EMv38kXrqRUNQFcK06hRpxTvVPgcMkKaKPJAyArjA=RzT-Q@mail.gmail.com>
References: <CAP_EMv38kXrqRUNQFcK06hRpxTvVPgcMkKaKPJAyArjA=RzT-Q@mail.gmail.com>
Message-ID: <CAEJPMR7ngx=WQr1Em+v94-nBjJLe2J=JaCqmhr=JM4Y=TbD_TQ@mail.gmail.com>

James.  Selection into the estimation sample is not supposed to be one of
the choices.  Your
choice set should not include "no trip."  On the other hand, if you want to
include NO TRIP
as one of the choices, then I suspect you should not be using the selection
model.  This is not
well defined for the selection model.  In theory, the selectivity treatment
is trying to isolate the
difference in trip choices between those "selected" and those "not
selected."  But, none of
your non-selected observations made a trip, so the comparison is not
possible.
/B. Greene

On Tue, Feb 15, 2022 at 4:21 PM James Giller via Limdep <
limdep at mailman.sydney.edu.au> wrote:

> Hi,
>
> My dependent variable is based on travel mode choice survey data, and it
> has a value for "make no trip at all" as well as values for various modes,
> car, transit, taxi, walking, etc. The mode is only observed if the survey
> participant would make a trip, which made me consider a sample selection
> model.
>
> My dependent variable Y has the following values:
> 0 = transit
> 1 = car
> 2 = carpool
> 3 = taxi
> 4 = walking
> 5 = other
> 6 = no trip
>
> I created another variable DOTRIP, which is 0 if Y = 6, and 1 otherwise.
>
> I then tried to estimate the following model
> PROBIT ; Lhs = DOTRIP ; Rhs = <explanators> ; Hold $
> MLOGIT ; Lhs = Y ; Rhs = <explanators> ; Selection ; Pts = 25; Halton $
>
> The above model on the data set as explained returned an error message that
> there is insufficient variation in the dependent variable during the MLOGIT
> part. I think this is because the value 6 (no trip) is possible for Y, but
> none of the observations would be selected in the selection step.
>
> Am I correct in specifying a model with sample selection in this case? Or
> should I take a different approach to account for the individuals first
> deciding whether or not to travel, then by which mode?
>
> Is there some way to rearrange my data so that the above model will work?
>
> Thank you.
> James.
> _______________________________________________
> Limdep site list
> Limdep at mailman.sydney.edu.au
> https://protect-au.mimecast.com/s/s3d_CmO5glujngZq3HG3TP8?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/8Q_ZCnx1jni7oPzV2CJLAzo?domain=people.stern.nyu.edu
Email: wgreene at stern.nyu.edu
Editor in Chief: Journal of Productivity Analysis
<https://protect-au.mimecast.com/s/NC3zCoV1kpfrzgMmRtVbiFz?domain=springer.com>
<https://protect-au.mimecast.com/s/NC3zCoV1kpfrzgMmRtVbiFz?domain=springer.com>
Editor in Chief: Foundations and Trends in Econometrics
Associate Editor: Economics Letters
Associate Editor: Journal of Business and Economic Statistics

From haab.1 at osu.edu  Thu Feb 17 01:57:33 2022
From: haab.1 at osu.edu (Haab, Timothy)
Date: Wed, 16 Feb 2022 14:57:33 +0000
Subject: [Limdep Nlogit List] Is a multinomial logit model with
	selection	applicable here?
In-Reply-To: <CAP_EMv38kXrqRUNQFcK06hRpxTvVPgcMkKaKPJAyArjA=RzT-Q@mail.gmail.com>
References: <CAP_EMv38kXrqRUNQFcK06hRpxTvVPgcMkKaKPJAyArjA=RzT-Q@mail.gmail.com>
Message-ID: <CH0PR01MB69722E7308B987CD109BD0D7DC359@CH0PR01MB6972.prod.exchangelabs.com>

James, looks like you will need a nested model (logit?) with the top level nest being trip/no trip, and then conditional on trip, mode.  

-----Original Message-----
From: limdep-bounces at mailman.sydney.edu.au <limdep-bounces at mailman.sydney.edu.au> On Behalf Of James Giller via Limdep
Sent: Tuesday, February 15, 2022 4:20 PM
To: limdep at mailman.sydney.edu.au
Subject: [Limdep Nlogit List] Is a multinomial logit model with selection applicable here?

Hi,

My dependent variable is based on travel mode choice survey data, and it has a value for "make no trip at all" as well as values for various modes, car, transit, taxi, walking, etc. The mode is only observed if the survey participant would make a trip, which made me consider a sample selection model.

My dependent variable Y has the following values:
0 = transit
1 = car
2 = carpool
3 = taxi
4 = walking
5 = other
6 = no trip

I created another variable DOTRIP, which is 0 if Y = 6, and 1 otherwise.

I then tried to estimate the following model PROBIT ; Lhs = DOTRIP ; Rhs = <explanators> ; Hold $ MLOGIT ; Lhs = Y ; Rhs = <explanators> ; Selection ; Pts = 25; Halton $

The above model on the data set as explained returned an error message that there is insufficient variation in the dependent variable during the MLOGIT part. I think this is because the value 6 (no trip) is possible for Y, but none of the observations would be selected in the selection step.

Am I correct in specifying a model with sample selection in this case? Or should I take a different approach to account for the individuals first deciding whether or not to travel, then by which mode?

Is there some way to rearrange my data so that the above model will work?

Thank you.
James.
_______________________________________________
Limdep site list
Limdep at mailman.sydney.edu.au
https://protect-au.mimecast.com/s/sNWFCr81nyt8vkpkxI7kLpu?domain=urldefense.com 


From jgiller at ucdavis.edu  Thu Feb 17 04:20:13 2022
From: jgiller at ucdavis.edu (James Giller)
Date: Wed, 16 Feb 2022 09:20:13 -0800
Subject: [Limdep Nlogit List] Is a multinomial logit model with
 selection applicable here?
In-Reply-To: <CH0PR01MB69722E7308B987CD109BD0D7DC359@CH0PR01MB6972.prod.exchangelabs.com>
References: <CAP_EMv38kXrqRUNQFcK06hRpxTvVPgcMkKaKPJAyArjA=RzT-Q@mail.gmail.com>
	<CH0PR01MB69722E7308B987CD109BD0D7DC359@CH0PR01MB6972.prod.exchangelabs.com>
Message-ID: <CAP_EMv0Ch0mn3GLFz3Vkpsm7XFsh4yrOwS=8Pw_sujBhPNOm7Q@mail.gmail.com>

Thank you William and Timothy. I will adjust my modeling approach.

On Wed, Feb 16, 2022, 6:57 AM Haab, Timothy <haab.1 at osu.edu> wrote:

> James, looks like you will need a nested model (logit?) with the top level
> nest being trip/no trip, and then conditional on trip, mode.
>
> -----Original Message-----
> From: limdep-bounces at mailman.sydney.edu.au <
> limdep-bounces at mailman.sydney.edu.au> On Behalf Of James Giller via Limdep
> Sent: Tuesday, February 15, 2022 4:20 PM
> To: limdep at mailman.sydney.edu.au
> Subject: [Limdep Nlogit List] Is a multinomial logit model with selection
> applicable here?
>
> Hi,
>
> My dependent variable is based on travel mode choice survey data, and it
> has a value for "make no trip at all" as well as values for various modes,
> car, transit, taxi, walking, etc. The mode is only observed if the survey
> participant would make a trip, which made me consider a sample selection
> model.
>
> My dependent variable Y has the following values:
> 0 = transit
> 1 = car
> 2 = carpool
> 3 = taxi
> 4 = walking
> 5 = other
> 6 = no trip
>
> I created another variable DOTRIP, which is 0 if Y = 6, and 1 otherwise.
>
> I then tried to estimate the following model PROBIT ; Lhs = DOTRIP ; Rhs =
> <explanators> ; Hold $ MLOGIT ; Lhs = Y ; Rhs = <explanators> ; Selection ;
> Pts = 25; Halton $
>
> The above model on the data set as explained returned an error message
> that there is insufficient variation in the dependent variable during the
> MLOGIT part. I think this is because the value 6 (no trip) is possible for
> Y, but none of the observations would be selected in the selection step.
>
> Am I correct in specifying a model with sample selection in this case? Or
> should I take a different approach to account for the individuals first
> deciding whether or not to travel, then by which mode?
>
> Is there some way to rearrange my data so that the above model will work?
>
> Thank you.
> James.
> _______________________________________________
> Limdep site list
> Limdep at mailman.sydney.edu.au
>
> https://protect-au.mimecast.com/s/Y_GeClx1Nji2ZVNVBIGzjDK?domain=urldefense.com
>
>

From ann.livingstone at sydney.edu.au  Thu Feb 17 09:04:26 2022
From: ann.livingstone at sydney.edu.au (Ann Livingstone)
Date: Wed, 16 Feb 2022 22:04:26 +0000
Subject: [Limdep Nlogit List] Recall: WTP/MRS calculation and reporting of
	95% CI
Message-ID: <SY2PR01MB2217193C5871C58E6903BB3FB9359@SY2PR01MB2217.ausprd01.prod.outlook.com>

Ann Livingstone would like to recall the message, "WTP/MRS calculation and reporting of 95% CI".


From doctorjasonong at gmail.com  Fri Feb 25 10:47:33 2022
From: doctorjasonong at gmail.com (Jason Ong)
Date: Fri, 25 Feb 2022 10:47:33 +1100
Subject: [Limdep Nlogit List] converting theta output from LCA to odds ratio
Message-ID: <CAG6D8PBTuGU83TovgwdvrrekNYiRU=7c-9wavzph+VuiHmd1nQ@mail.gmail.com>

hi,

In a latent class analysis where I put in sociodemographic characteristics
to see if these are more likely in a class, the NLOGIT output presents
these as theta
If I want to convert these to odds ratios, can I exponentiate these theta
values?

Thanks for confirming.


*Jason Ong*
Twitter: @DrJasonJOng
PhD, MMed, MBBS, FAChSHM, FRACGP

Sexual Health Physician, Melbourne Sexual Health Centre (MSHC)
Head, HIV/STI Economics and Health Preference Research (MSHC)

Associate Professor (Hon), London School of Hygiene and Tropical Medicine,
UK
Central Clinical School, Monash University, Australia
Melbourne School of Population and Global Health, University of Melbourne,
Australia
Australian NHRMC Emerging Leader Fellow

Associate Editor, Sexually Transmitted Infections
Editorial Board Member, Sexually Transmitted Diseases
Special Issues Editor, Sexual Health
Board Director, ASHM (https://protect-au.mimecast.com/s/Rbi4C4QOPEiBXYZmkhObBwe?domain=ashm.org.au)
Board Director, AFAO (https://protect-au.mimecast.com/s/ttLXC5QPXJiZBgyxNUOCrjD?domain=afao.org.au)
Executive Member, ASHRA (
https://protect-au.mimecast.com/s/tF_sC6XQ4Lfr5VR1wcmdeJH?domain=sexualreproductivehealthalliance.org.au
My publications: https://protect-au.mimecast.com/s/hX6KC71R2NTADErQ7INZAwR?domain=lshtm.ac.uk