From brett.smith at uwa.edu.au Mon Dec 2 12:12:26 2019 From: brett.smith at uwa.edu.au (Brett Smith) Date: Mon, 2 Dec 2019 01:12:26 +0000 Subject: [Limdep Nlogit List] manuals Message-ID: Hi all, I recently moved offices and decided to discard my old hardcopy manuals thinking I could obtain newer online pdf's. However, despite visiting the website a couple of times I cannot find the appropriate link. Can anyone send the link to pdf version of LimDep manuals? Thank you in advance Brett Brett Smith | PhD Department of Management & Organisations Director of Masters of Business Information and Logistics Management Research Fellow, Centre for Business Analytics Research Associate, Planning and Transport Research Center (PATREC) UWA Business School | M261, Perth WA 6009 Australia T +61 8 6488 3979 | E Brett.Smith at uwa.edu.au From wgreene at stern.nyu.edu Mon Dec 2 23:14:30 2019 From: wgreene at stern.nyu.edu (William Greene) Date: Mon, 2 Dec 2019 07:14:30 -0500 Subject: [Limdep Nlogit List] manuals In-Reply-To: References: Message-ID: Brett. There is a copy of the manual on your computer in the C:\LIMDEP11\Documentation folder. /Bill Greene On Sun, Dec 1, 2019 at 8:12 PM Brett Smith wrote: > Hi all, > > I recently moved offices and decided to discard my old hardcopy manuals > thinking I could obtain newer online pdf's. However, despite visiting the > website a couple of times I cannot find the appropriate link. Can anyone > send the link to pdf version of LimDep manuals? > > Thank you in advance > > Brett > > Brett Smith | PhD > Department of Management & Organisations > Director of Masters of Business Information and Logistics Management > Research Fellow, Centre for Business Analytics > Research Associate, Planning and Transport Research Center (PATREC) > UWA Business School | M261, Perth WA 6009 Australia > T +61 8 6488 3979 | E Brett.Smith at uwa.edu.au > > > _______________________________________________ > Limdep site list > Limdep at mailman.sydney.edu.au > http://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/oEvxCANZvPik9kpYSGjArL?domain=people.stern.nyu.edu Email: wgreene at stern.nyu.edu Ph. +1.646.596.3296 Editor in Chief: Journal of Productivity Analysis Editor in Chief: Foundations and Trends in Econometrics Associate Editor: Economics Letters Associate Editor: Journal of Business and Economic Statistics From cvmh.kiel at gmail.com Wed Dec 4 06:59:20 2019 From: cvmh.kiel at gmail.com (Cordula H) Date: Tue, 3 Dec 2019 20:59:20 +0100 Subject: [Limdep Nlogit List] Cluster-robust standard errors (panel-adjusted) for conditional logit model in Nlogit? In-Reply-To: References: Message-ID: Dear Professor Greene, Thank you for your reply! I have tried this as you suggested, but unfortunately it did not change anything in the results. I use NLOGIT 5, not 6, if this is relevant. Kind regards Cordula On Wed, Nov 27, 2019 at 8:36 PM William Greene via Limdep < limdep at mailman.sydney.edu.au> wrote: > Cordula. Assuming you are using version 6.0 of NLOGIT, the appropriate > spec. > is ;Cluster = number of choice tasks. For you, that would be ;Cluster = 8. > Regards, > Bill Greene > > On Sun, Nov 17, 2019 at 6:44 AM Cordula H via Limdep < > limdep at mailman.sydney.edu.au> wrote: > > > Dear all, > > > > > > I am new to this list and also new to Nlogit. ? > > > > I have some basic experience with choice analysis, but I have only used > > Stata so far. > > Now I am supposed to use Nlogit and I have some issues figuring out the > > right commands, although I have tried to find answers in the Nlogit > > documentation. > > > > > > I have conducted a pilot study of the choice experiment to get estimates > > that I can use as priors for input into Ngene to generate an efficient > > design. Each of the respondents got 8 choice sets out of a blocked 3x8 > > design. Each choice tasks consists of three alternatives. > > > > Now I tried to estimate a conditional logit model. > > > > > > First, I used Stata. I estimated the model with cluster-robust standard > > errors, because I thought that it makes sense to account for the panel > > data, so that it is recognized that one respondent is associated with > > several observations. > > > > However, I could not figure out yet how to do this in Nlogit. I > estimated a > > conditional logit model, which provides the same results as the basic > > conditional logit model in Stata, without accounting for the panel data / > > cluster-robust standard errors. I found some commands like "cluster" or > > "robust" for Nlogit, but somehow this does not really seem to work. > > Probably I am not using these commands correctly. (For instance, I tried > > "cluster = ID", with ID being the individual ID of each respondent. The > > results are different than the results of the basic model, but not the > same > > as in the Stata version, and in the output the number of respondents does > > not seem to be recognized. > > > > > > Could anybody please help me in this regard? > > > > > > Thanks a lot in advance! > > _______________________________________________ > > Limdep site list > > Limdep at mailman.sydney.edu.au > > http://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/HIssC71ZgLtDZgvri8QCMq?domain=people.stern.nyu.edu > Email: wgreene at stern.nyu.edu > Ph. +1.646.596.3296 > Editor in Chief: Journal of Productivity Analysis > Editor in Chief: Foundations and Trends in Econometrics > Associate Editor: Economics Letters > Associate Editor: Journal of Business and Economic Statistics > _______________________________________________ > Limdep site list > Limdep at mailman.sydney.edu.au > http://limdep.itls.usyd.edu.au > > From wgreene at stern.nyu.edu Fri Dec 6 05:33:13 2019 From: wgreene at stern.nyu.edu (William Greene) Date: Thu, 5 Dec 2019 13:33:13 -0500 Subject: [Limdep Nlogit List] simulation of ordered logit probabilities after inclusion of random parameters In-Reply-To: References: Message-ID: Derek. This command does not fit a random parameters model. In order to fit an RPM you must include the specification of which parameters are random and what the distributions are. Although the command did not work to fit an RPM, the presence of the ;RPM switch in the command turns off the saving of the results for the subsequent PARTIALS and SIMULATE commands, which partly explains the diagnostic about SIMULATE. Before we turn to that, it looks like the ;RST specification specifies far more parameters than there are in the model. Finally, it is unclear with the RTM (random thresholds) model is doing in here. I could not figure out what is going on with ;RST. The SIMULATE command is not set up to work with a random parameters model. What you can do to get the mean estimates of simulation or partial is, after fitting this model, use the betas and mus as starting values in an ordinary ORDERED command and include ;MAXIT=0. This will store the model in a way that can be used with PARTIALS and SIMULATE. /Bill Greene On Mon, Nov 18, 2019 at 8:07 AM Derek Monsuur wrote: > Dear sir/madam, > I have a question with regards to the simulation of probabilities after > estimating an ordered logit model with random parameters. After estimation, > I like to simulate probabilities for several scenario's, in which the > parameter 'MIN' is varied over several levels, e.g. from 0 to 80. The model > looks like this: > ORDERED; > LHS=SAT; > RHS = one, FEM, FREQ, MIN, STAND, INFO, IC, BCANCEL, CANCEL, AGE1, AGE3, > MG3, COMB, PERB, LEI, COMN, RED; > > RST=b1,b2,b3,b4,b5,b6,b7,b8,b9,b10,b11,b12,b13,b14,b15,b16,b17,b18,b19,b20,b21,b22,b23,0,0,0,b27,0,b29,0,0,0,0,0,0,0,0,0,0,0; > RTM; > RPM; > halton;pts=20; > alg=bfgs; > logit;partials$ > And after model estimation I use the following command: > Simulation;Scenario:&MIN=0(5)80;Outcome=*$ > This command works perfectly fine with a basic ordered logit model, > however, when I use this command after estimating the model with random > thresholds and random parameters (for 'MIN' and 'INFO'), I get the > following error message: > "Previous model did not store model setup for PARTIALS. > Check model setup and output for errors. Setup does > not occur if model command contains Expand(variable)." > To me it seems that I somehow need to store the individual specific > parameter means, and according to the manual this can be done with the > command ;Parameters. If I include this command however, NLOGIT crashes > immediately, without giving me the opportunity to inspect the error > message. Or is it the case that this functionality is simply not available > for ordered logit with random parameters? > With kind regards, > Derek Monsuur > > _______________________________________________ > Limdep site list > Limdep at mailman.sydney.edu.au > http://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/jfzrCyoNVrcxE61mSZPODq?domain=people.stern.nyu.edu Email: wgreene at stern.nyu.edu Ph. +1.646.596.3296 Editor in Chief: Journal of Productivity Analysis Editor in Chief: Foundations and Trends in Econometrics Associate Editor: Economics Letters Associate Editor: Journal of Business and Economic Statistics From wgreene at stern.nyu.edu Fri Dec 6 09:30:21 2019 From: wgreene at stern.nyu.edu (William Greene) Date: Thu, 5 Dec 2019 17:30:21 -0500 Subject: [Limdep Nlogit List] About Heteroskedastic Ordered Probit In-Reply-To: References: Message-ID: David. I believe all you have to do with this command is change ;RH2 to ;HFN in the second command. Regards Bill Greene On Fri, Nov 8, 2019 at 1:28 PM DAVID BOTO GARCIA wrote: > Hello, > > I am using LIMDEP 11 and I am trying to estimate a Heteroskedastic Ordered > Probit, where both the mean and the variance (multiplicative) are modelled > as a function of covariates. I used the commands proposed in the > Econometric Modelling Guide. However, when I run the model LIMDEP estimates > only the Homoskedastic Model. I am not able to estimate the heteroskedastic > version even through I specifically asked for it. I wonder whether my code > is wrong or there might be another problem. > > namelist; > x=hom,age,tracp,est,ama,par,jub,ext,res,vac,trab,alone,pare,los,first,cy,orcos, > > orint,nuacom,mcal,mcon,mexp,mgus,mrec,msit,mtran,mcer,mpre,aldes,mpen,pcom,gasto,s3,s4,s5, > t3,y06,y07,y08,y09,y10,y11,y12,y13,y14,y15,y16; > z=age,hom,first,pcom,cy,orcos,orint,s3,s4,s5,mcon,mgus,msit,mpre $ > > ?BASELINE ORDERED MODEL > ORDERED ; Lhs = y ; Rhs = one,x $ > > ?HETEROSKEDASTIC ORDERED MODEL > ORDERED ; Lhs = y ; Rhs = x; Heteroscedastic; > Rh2=age,hom,first,pcom,cy,orcos,orint,s3,s4,s5,mcon,mgus,msit,mpre$ > > Regards, > David > > _______________________________________________ > Limdep site list > Limdep at mailman.sydney.edu.au > http://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/QhovC6X13Rt5xLykSpHxtH?domain=people.stern.nyu.edu Email: wgreene at stern.nyu.edu Ph. +1.646.596.3296 Editor in Chief: Journal of Productivity Analysis Editor in Chief: Foundations and Trends in Econometrics Associate Editor: Economics Letters Associate Editor: Journal of Business and Economic Statistics From ashu.kedia at pg.canterbury.ac.nz Tue Dec 10 11:08:22 2019 From: ashu.kedia at pg.canterbury.ac.nz (Ashu Kedia) Date: Tue, 10 Dec 2019 00:08:22 +0000 Subject: [Limdep Nlogit List] Request to help with estimating marginal effects In-Reply-To: <016F0B45C2610A44B0B3F76A466532E0011CE2F8@UCEXMBX03-I.canterbury.ac.nz> References: <016F0B45C2610A44B0B3F76A466532E0011CE2F8@UCEXMBX03-I.canterbury.ac.nz> Message-ID: <016F0B45C2610A44B0B3F76A466532E0011CE38A@UCEXMBX03-I.canterbury.ac.nz> Hi Professor William Greene, I hope you are doing well. I am Ashu Kedia, a PhD student in transportation engineering at the University of Canterbury, New Zealand. I am having a few doubts regarding marginal effects calculation. It will be a great help if you could clarify my doubts. 1. I am using Nlogit 5 version to carry out estimations, but have been following the Hensher et al. (2005) book to understand the process, which might have used Nlogit 3 version. The book says (on pg. 398 and 399) that the Marginal effects as reported are ?true marginal effects multiplied by 100?. So, we need to divide the reported values by 100 to obtain the true marginal effect value. Now, I presume that this is applicable to that particular version of the software (i.e. Nlogit 3 or 4). However, I am using Nlogit 5 version, and it is not clear to me whether the above convention is applicable to the marginal effects results given by the Nlogit 5 version as well? I am asking this because the marginal effects obtained in my study are already very small in magnitude. For example: they vary between 0.08 and 1.06. If I divide such small values by 100, it will be exceptionally small. Is this not uncommon? Could you please advise on this? 2. The examples of marginal effects calculations given in the book, Hensher et al. (2005), are for the MNL model specification. Do the syntax remains the same if marginal effects are calculated for a mixed logit model? 3. Hensher et al. (2005) book has shown two different methods to calculate marginal effects for continuous-level data and categorically coded data. Details of attributes and levels adopted in my study are as follows: Alternatives: 2 Attributes: continuous (e.g. cost, time, etc.) ? 4 attributes with 3 levels in each; categorical (i.e. qualitative attributes; e.g. type of a facility, such as residential or commercial) ? 3 attributes with 2 levels in each. Model type: Mixed logit model with heterogeneity in mean Marginal effect estimation method: ;Pwt (probability weighted average) Question : I have estimated the marginal effects of the continuous level attributes, using the method given on page no. 397 (but applying changes to the syntax that are required to estimate a mixed logit model). To estimate the marginal effects for the categorically coded variables, the book shows the ?Simulation? method to be used. However, when we use these two different methods separately to estimate marginal effects (i.e. once for continuous variables and then for categorical variables), two sets of coefficient values (i.e. parameter estimates for the model) are obtained. But we need only one set of coefficients for the estimated model. Which one of those two sets of estimates are to be adopted for the final reporting? It will be a great help if you clarify these questions. I am happy to provide more details, if needed. Kind regards Ashu Ashu Kedia PhD Student, Department of Civil and Natural Resources Engineering, University of Canterbury, Christchurch 8140 New Zealand ________________________________ From: Ashu Kedia Sent: Tuesday, December 10, 2019 12:57 PM To: limdep at limdep.itls.usyd.edu.au Subject: Request to help with estimating marginal effects Hi Professor William Greene, I hope you are doing well. I am Ashu Kedia, a PhD student in transportation engineering at the University of Canterbury, New Zealand. I am having a few doubts regarding marginal effects calculation. It will be a great help if you could clarify my doubts. 1. I am using Nlogit 5 version to carry out estimations, but have been following the Hensher et al. (2005) book to understand the process, which might have used Nlogit 3 version. The book says (on pg. 398 and 399) that the Marginal effects as reported are ?true marginal effects multiplied by 100?. So, we need to divide the reported values by 100 to obtain the true marginal effect value. Now, I presume that this is applicable to that particular version of the software (i.e. Nlogit 3 or 4). However, I am using Nlogit 5 version, and it is not clear to me whether the above convention is applicable to the marginal effects results given by the Nlogit 5 version as well? I am asking this because the marginal effects obtained in my study are already very small in magnitude. For example: they vary between 0.08 and 1.06. If I divide such small values by 100, it will be exceptionally small. Is this not uncommon? Could you please advise on this? 2. The examples of marginal effects calculations given in the book, Hensher et al. (2005), are for the MNL model specification. Do the syntax remains the same if marginal effects are calculated for a mixed logit model? 3. Hensher et al. (2005) book has shown two different methods to calculate marginal effects for continuous-level data and categorically coded data. Details of attributes and levels adopted in my study are as follows: Alternatives: 2 Attributes: continuous (e.g. cost, time, etc.) ? 4 attributes with 3 levels in each; categorical (i.e. qualitative attributes; e.g. type of a facility, such as residential or commercial) ? 3 attributes with 2 levels in each. Model type: Mixed logit model with heterogeneity in mean Marginal effect estimation method: ;Pwt (probability weighted average) Question : I have estimated the marginal effects of the continuous level attributes, using the method given on page no. 397 (but applying changes to the syntax that are required to estimate a mixed logit model). To estimate the marginal effects for the categorically coded variables, the book shows the ?Simulation? method to be used. However, when we use these two different methods separately to estimate marginal effects (i.e. once for continuous variables and then for categorical variables), two sets of coefficient values (i.e. parameter estimates for the model) are obtained. But we need only one set of coefficients for the estimated model. Which one of those two sets of estimates are to be adopted for the final reporting? It will be a great help if you clarify these questions. I am happy to provide more details, if needed. Kind regards Ashu From doctorjasonong at gmail.com Tue Dec 10 16:41:05 2019 From: doctorjasonong at gmail.com (Jason Ong) Date: Tue, 10 Dec 2019 16:41:05 +1100 Subject: [Limdep Nlogit List] Error 1097: Unusable LHS variable in choice model setup Message-ID: I am doing a preliminary analysis on a DCE and am trying to run a MNL model using NLOGIT 6 But I came across this error - "Error 1097: Unusable LHS variable in choice model setup" I am not sure what I am doing wrong with my code (data set also attached as *.csv file) NLOGIT ; Lhs = ChoiceV, cset, altij ; Choices = A,B,C ; Model: U(A,B)=prep1*prep1+prep2*prep2+condom*condom+sex1*sex1+sex2*sex2+test1*test1+test2*test2+pep*pep/ U(C)=neither $ thanks for your help. *Jason Ong* From gk3saini at uwaterloo.ca Tue Dec 10 17:42:39 2019 From: gk3saini at uwaterloo.ca (Gursimran Kaur Saini) Date: Tue, 10 Dec 2019 06:42:39 +0000 Subject: [Limdep Nlogit List] (no subject) Message-ID: <9FA880AA-ADA3-44B0-AABE-FAABE575A4C6@uwaterloo.ca> Hello, I am trying to develop a mode choice model using a nested Logit model using NLOGIT 6. I am running into the following errors: - I am unable to load my data file onto NLOGIT (with about 250,000 observations), when I create another file with a part of the data, I was able to import it, but not the complete file. - Number of choices not equal to the number of alternatives? I?m not sure what I am doing wrong. Here?s the code I am using Nlogit ;Lhs=CHOICE ;Choices=CarD,CarP,Bus,Walk,Bike,MSM ;Tree=Trips{Private[Car(CarD,CarP),MSM],Public[Bus],NMT[Walk,Bike]} ;ias=Carp ;Show Tree ;start=logit ;Maxit=100 ;Model: U(CarD) = a_CarD+tt*TT+vo*VOwn+emp*EMPD+Inc*Income+popn*PopD / U(CarP) = b_CarP+tt*TT+vo*VOwn+emp*EMPD+Inc*Income+popn*PopD / U(Bus) = c_Bus+tt*TT+vo*VOwn+emp*EMPD+Inc*Income+popn*PopD / U(Walk) = d_Walk+tt*TT+vo*VOwn+emp*EMPD+Inc*Income+popn*PopD / U(Bike) = e_Bike+tt*TT+vo*VOwn+emp*EMPD+Inc*Income+popn*PopD / U(MSM) = e_MSM+tt*TT+vo*VOwn+emp*EMPD+Inc*Income+popn*PopD $ Please help me resolve. thank you, Gursimran From m.umair.k104 at gmail.com Wed Dec 11 15:02:41 2019 From: m.umair.k104 at gmail.com (Muhammad Umair Khan) Date: Wed, 11 Dec 2019 15:02:41 +1100 Subject: [Limdep Nlogit List] Simulation Scenario Message-ID: Dear All, I am a newbie and just started using DCE and NLOGIT. I have three alternatives; Medication A, Medication B, No medication. One of my attributes is side-effects (SE) (0= no SE, 1= Mild SE, 2= Moderate SE, 3= severe SE). I am trying to create a simulation scenario by checking the probability of taking medication if the severity of side-effects change from 1 to 3. I am using the following command: ;simulate ;scenario:sidemi(*)=[+]2 I am assuming that by using [+]2 NLOGIT would assume that I want to go two levels up from 1, that is, from 1 to 3 (mild to severe). Am I correct here? If no, what should be the best way? Furthermore the output I am getting is, for example, the probability of people taking medication would reduce by 10%. I am interesting in finding out if this 10% is statistically significant? Any command through which we can get additional statistics such as p-value? Thank you! Umair From wgreene at stern.nyu.edu Thu Dec 12 04:56:37 2019 From: wgreene at stern.nyu.edu (William Greene) Date: Wed, 11 Dec 2019 12:56:37 -0500 Subject: [Limdep Nlogit List] Simulation Scenario In-Reply-To: References: Message-ID: Umair. Yes, it will change a 1 to 3. It will also change a 2 to a 4. Regards Bill Greene On Tue, Dec 10, 2019 at 11:03 PM Muhammad Umair Khan via Limdep < limdep at mailman.sydney.edu.au> wrote: > Dear All, > > I am a newbie and just started using DCE and NLOGIT. I have three > alternatives; Medication A, Medication B, No medication. One of my > attributes is side-effects (SE) (0= no SE, 1= Mild SE, 2= Moderate SE, 3= > severe SE). I am trying to create a simulation scenario by checking the > probability of taking medication if the severity of side-effects change > from 1 to 3. I am using the following command: > > ;simulate > ;scenario:sidemi(*)=[+]2 > > I am assuming that by using [+]2 NLOGIT would assume that I want to go two > levels up from 1, that is, from 1 to 3 (mild to severe). Am I correct here? > If no, what should be the best way? Furthermore the output I am getting is, > for example, the probability of people taking medication would reduce by > 10%. I am interesting in finding out if this 10% is statistically > significant? Any command through which we can get additional statistics > such as p-value? > > Thank you! > > Umair > _______________________________________________ > Limdep site list > Limdep at mailman.sydney.edu.au > http://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/6iWjCgZowLHMv1GnSNxTZw?domain=people.stern.nyu.edu Email: wgreene at stern.nyu.edu Ph. +1.646.596.3296 Editor in Chief: Journal of Productivity Analysis Editor in Chief: Foundations and Trends in Econometrics Associate Editor: Economics Letters Associate Editor: Journal of Business and Economic Statistics From wgreene at stern.nyu.edu Mon Dec 16 08:57:33 2019 From: wgreene at stern.nyu.edu (William Greene) Date: Sun, 15 Dec 2019 16:57:33 -0500 Subject: [Limdep Nlogit List] Error 1097: Unusable LHS variable in choice model setup In-Reply-To: References: Message-ID: Jason. The variable CHOICEV in your data set is not coded correctly for the choice model. It is supposed to be zeros and ones, throughout. It is unclear what you are using, but the error 14 indicates that every observation is badly coded. /Bill Greene On Tue, Dec 10, 2019 at 12:41 AM Jason Ong via Limdep < limdep at mailman.sydney.edu.au> wrote: > I am doing a preliminary analysis on a DCE and am trying to run a MNL model > using NLOGIT 6 > But I came across this error - "Error 1097: Unusable LHS variable in > choice model setup" > I am not sure what I am doing wrong with my code (data set also attached as > *.csv file) > > NLOGIT > ; Lhs = ChoiceV, cset, altij > ; Choices = A,B,C > ; Model: > > U(A,B)=prep1*prep1+prep2*prep2+condom*condom+sex1*sex1+sex2*sex2+test1*test1+test2*test2+pep*pep/ > U(C)=neither > $ > > thanks for your help. > > *Jason Ong* > _______________________________________________ > Limdep site list > Limdep at mailman.sydney.edu.au > http://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/ajAoC6X13Rt5krmgIpU_5O?domain=people.stern.nyu.edu Email: wgreene at stern.nyu.edu Ph. +1.646.596.3296 Editor in Chief: Journal of Productivity Analysis Editor in Chief: Foundations and Trends in Econometrics Associate Editor: Economics Letters Associate Editor: Journal of Business and Economic Statistics From wgreene at stern.nyu.edu Mon Dec 16 09:12:51 2019 From: wgreene at stern.nyu.edu (William Greene) Date: Sun, 15 Dec 2019 17:12:51 -0500 Subject: [Limdep Nlogit List] (no subject) In-Reply-To: <9FA880AA-ADA3-44B0-AABE-FAABE575A4C6@uwaterloo.ca> References: <9FA880AA-ADA3-44B0-AABE-FAABE575A4C6@uwaterloo.ca> Message-ID: Gursimran. You can expand the memory space used for data by using Tools->Options->Projects. Expand the number of cells, then exit and restart the program. Your model is overspecified. You can't have constants for every alternative. At least one must be omitted. It's not clear what function the ;IAS=CarP is playing in your model, but you probably don't want it in there. It is used to set up a Hausman test, but you should only use it for the MNL. Finally, I'm not certain your specification is rich enough to support a three level model like that. You might try starting with something very simple and building it up a bit at a time. /Bill Greene On Tue, Dec 10, 2019 at 1:43 AM Gursimran Kaur Saini via Limdep < limdep at mailman.sydney.edu.au> wrote: > Hello, > > I am trying to develop a mode choice model using a nested Logit model > using NLOGIT 6. > I am running into the following errors: > - I am unable to load my data file onto NLOGIT (with about 250,000 > observations), when I create another file with a part of the data, I was > able to import it, but not the complete file. > - Number of choices not equal to the number of alternatives? > > I?m not sure what I am doing wrong. > Here?s the code I am using > > Nlogit > ;Lhs=CHOICE > > ;Choices=CarD,CarP,Bus,Walk,Bike,MSM > ;Tree=Trips{Private[Car(CarD,CarP),MSM],Public[Bus],NMT[Walk,Bike]} > ;ias=Carp > ;Show Tree > ;start=logit > ;Maxit=100 > ;Model: > U(CarD) = a_CarD+tt*TT+vo*VOwn+emp*EMPD+Inc*Income+popn*PopD / > U(CarP) = b_CarP+tt*TT+vo*VOwn+emp*EMPD+Inc*Income+popn*PopD / > U(Bus) = c_Bus+tt*TT+vo*VOwn+emp*EMPD+Inc*Income+popn*PopD / > U(Walk) = d_Walk+tt*TT+vo*VOwn+emp*EMPD+Inc*Income+popn*PopD / > U(Bike) = e_Bike+tt*TT+vo*VOwn+emp*EMPD+Inc*Income+popn*PopD / > U(MSM) = e_MSM+tt*TT+vo*VOwn+emp*EMPD+Inc*Income+popn*PopD $ > > Please help me resolve. > thank you, > Gursimran > _______________________________________________ > Limdep site list > Limdep at mailman.sydney.edu.au > http://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/aHStCL7rK8t6qk80CBl46-?domain=people.stern.nyu.edu Email: wgreene at stern.nyu.edu Ph. +1.646.596.3296 Editor in Chief: Journal of Productivity Analysis Editor in Chief: Foundations and Trends in Econometrics Associate Editor: Economics Letters Associate Editor: Journal of Business and Economic Statistics From m.umair.k104 at gmail.com Tue Dec 17 15:33:35 2019 From: m.umair.k104 at gmail.com (Muhammad Umair Khan) Date: Tue, 17 Dec 2019 15:33:35 +1100 Subject: [Limdep Nlogit List] Simulation Scenario In-Reply-To: References: Message-ID: Dear Bill, Thank you very much for your reply. Can I also use the [+]1 scenario command in case of dummy-coded variables? For example, I have a dummy-coded variable, mild side-effects, coded as 0 (no side-effects) and 1 (mild side-effects). Can I simulate a scenario "=[+]1 interpreting that there would be one level increase in the severity of mild side-effects, that is, the severity would increase from to mild to, let's say, moderate? Thank you for your time. Umair On Thu, Dec 12, 2019 at 4:57 AM William Greene via Limdep < limdep at mailman.sydney.edu.au> wrote: > Umair. Yes, it will change a 1 to 3. It will also change a 2 to a 4. > Regards > Bill Greene > > On Tue, Dec 10, 2019 at 11:03 PM Muhammad Umair Khan via Limdep < > limdep at mailman.sydney.edu.au> wrote: > > > Dear All, > > > > I am a newbie and just started using DCE and NLOGIT. I have three > > alternatives; Medication A, Medication B, No medication. One of my > > attributes is side-effects (SE) (0= no SE, 1= Mild SE, 2= Moderate SE, 3= > > severe SE). I am trying to create a simulation scenario by checking the > > probability of taking medication if the severity of side-effects change > > from 1 to 3. I am using the following command: > > > > ;simulate > > ;scenario:sidemi(*)=[+]2 > > > > I am assuming that by using [+]2 NLOGIT would assume that I want to go > two > > levels up from 1, that is, from 1 to 3 (mild to severe). Am I correct > here? > > If no, what should be the best way? Furthermore the output I am getting > is, > > for example, the probability of people taking medication would reduce by > > 10%. I am interesting in finding out if this 10% is statistically > > significant? Any command through which we can get additional statistics > > such as p-value? > > > > Thank you! > > > > Umair > > _______________________________________________ > > Limdep site list > > Limdep at mailman.sydney.edu.au > > http://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/1BsPCoVzGQigXmgJi1TdeZ?domain=people.stern.nyu.edu > Email: wgreene at stern.nyu.edu > Ph. +1.646.596.3296 > Editor in Chief: Journal of Productivity Analysis > Editor in Chief: Foundations and Trends in Econometrics > Associate Editor: Economics Letters > Associate Editor: Journal of Business and Economic Statistics > _______________________________________________ > Limdep site list > Limdep at mailman.sydney.edu.au > http://limdep.itls.usyd.edu.au > > From david.hensher at sydney.edu.au Tue Dec 17 15:38:32 2019 From: david.hensher at sydney.edu.au (David Hensher) Date: Tue, 17 Dec 2019 15:38:32 +1100 Subject: [Limdep Nlogit List] Simulation Scenario In-Reply-To: References: Message-ID: <5DF85BC8.2010207@sydney.edu.au> I think this will result in 0 and 1 becoming 1 and 2 so instead use =1 David Hensher On 17/12/2019 3:33 PM, Muhammad Umair Khan via Limdep wrote: > Dear Bill, > > Thank you very much for your reply. Can I also use the [+]1 scenario > command in case of dummy-coded variables? For example, I have a dummy-coded > variable, mild side-effects, coded as 0 (no side-effects) and 1 (mild > side-effects). Can I simulate a scenario "=[+]1 interpreting that there > would be one level increase in the severity of mild side-effects, that is, > the severity would increase from to mild to, let's say, moderate? > > Thank you for your time. > > Umair > > On Thu, Dec 12, 2019 at 4:57 AM William Greene via Limdep< > limdep at mailman.sydney.edu.au> wrote: > > >> Umair. Yes, it will change a 1 to 3. It will also change a 2 to a 4. >> Regards >> Bill Greene >> >> On Tue, Dec 10, 2019 at 11:03 PM Muhammad Umair Khan via Limdep< >> limdep at mailman.sydney.edu.au> wrote: >> >> >>> Dear All, >>> >>> I am a newbie and just started using DCE and NLOGIT. I have three >>> alternatives; Medication A, Medication B, No medication. One of my >>> attributes is side-effects (SE) (0= no SE, 1= Mild SE, 2= Moderate SE, 3= >>> severe SE). I am trying to create a simulation scenario by checking the >>> probability of taking medication if the severity of side-effects change >>> from 1 to 3. I am using the following command: >>> >>> ;simulate >>> ;scenario:sidemi(*)=[+]2 >>> >>> I am assuming that by using [+]2 NLOGIT would assume that I want to go >>> >> two >> >>> levels up from 1, that is, from 1 to 3 (mild to severe). Am I correct >>> >> here? >> >>> If no, what should be the best way? Furthermore the output I am getting >>> >> is, >> >>> for example, the probability of people taking medication would reduce by >>> 10%. I am interesting in finding out if this 10% is statistically >>> significant? Any command through which we can get additional statistics >>> such as p-value? >>> >>> Thank you! >>> >>> Umair >>> _______________________________________________ >>> Limdep site list >>> Limdep at mailman.sydney.edu.au >>> http://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/cq8aCXLKZoi8nEE5u658JA?domain=people.stern.nyu.edu >> Email: wgreene at stern.nyu.edu >> Ph. +1.646.596.3296 >> Editor in Chief: Journal of Productivity Analysis >> Editor in Chief: Foundations and Trends in Econometrics >> Associate Editor: Economics Letters >> Associate Editor: Journal of Business and Economic Statistics >> _______________________________________________ >> Limdep site list >> Limdep at mailman.sydney.edu.au >> http://limdep.itls.usyd.edu.au >> >> >> > _______________________________________________ > Limdep site list > Limdep at mailman.sydney.edu.au > http://limdep.itls.usyd.edu.au > > -- DAVID HENSHER FASSA, PhD| Professor and Founding Director Institute of Transport and Logistics Studies | The University of Sydney Business School THE UNIVERSITY OF SYDNEY Rm 201, Building H73| The University of Sydney | NSW | 2006 Street Address: 378 Abercrombie St, Darlington NSW 2008 T +61 2 9114 1871 | F +61 2 9114 1863 | M +61 418 433 057 E David.Hensher at sydney.edu.au | W https://protect-au.mimecast.com/s/36ZECYWL1vin3VV1CGXFtP?domain=sydney.edu.au |W https://protect-au.mimecast.com/s/Q_nnCZYM2VFkMKK4IKPrXf?domain=sydney.edu.au Celebrating 25 years of ITLS: 1991-2016 https://protect-au.mimecast.com/s/7JgJC1WZXriPpJJzSX0Tj9?domain=youtu.be ERA Rank 5 (Transportation and Freight Services) Co-Founder of the International Conference Series on Competition and Ownership of Land Passenger Transport (The 'Thredbo' Series) https://protect-au.mimecast.com/s/m4MWC2xZYvCBk55lT9R0St?domain=thredbo-conference-series.org https://www.linkedin.com/company/28450714 Second edition of Applied Choice Analysis now available at https://protect-au.mimecast.com/s/AoBZC3Q8Z2FymDDAhvuoGK?domain=cambridge.org Nlogit is the most popular software for choice modellers. See https://protect-au.mimecast.com/s/CnY4C4QZ1RFDJ440uj0ROq?domain=limdep.com See Master of Transport beginning 2020: * Master of Transport: https://sydney.edu.au/courses/courses/pc/master-of-transport.html * Graduate Certificate in Transport: https://sydney.edu.au/courses/courses/pc/graduate-certificate-in-transport.html * Graduate Diploma in Transport: https://sydney.edu.au/courses/courses/pc/graduate-diploma-in-transport.html CRICOS 00026A This email plus any attachments to it are confidential. Any unauthorised use is strictly prohibited. If you receive this email in error, please delete it and any attachments. Please think of our environment and only print this e-mail if necessary. From wgreene at stern.nyu.edu Wed Dec 18 02:12:02 2019 From: wgreene at stern.nyu.edu (William Greene) Date: Tue, 17 Dec 2019 10:12:02 -0500 Subject: [Limdep Nlogit List] Request to help with estimating marginal effects In-Reply-To: <016F0B45C2610A44B0B3F76A466532E0011CE38A@UCEXMBX03-I.canterbury.ac.nz> References: <016F0B45C2610A44B0B3F76A466532E0011CE2F8@UCEXMBX03-I.canterbury.ac.nz> <016F0B45C2610A44B0B3F76A466532E0011CE38A@UCEXMBX03-I.canterbury.ac.nz> Message-ID: Ashu: 1. The calculation of the elasticities has always been the same. 2. Yes. But, keep in mind, you are computing partial effects for an average coefficient, since the coefficients are random. 3. I don't quite understand this question. But, you only need to fit the model once, then you can follow with more than one SIMULATE command. Regards, Bill Greene On Mon, Dec 9, 2019 at 7:08 PM Ashu Kedia wrote: > Hi Professor William Greene, > I hope you are doing well. > > I am Ashu Kedia, a PhD student in transportation engineering at the > University of Canterbury, New Zealand. > > I am having a few doubts regarding marginal effects calculation. It will > be a great help if you could clarify my doubts. > > > 1. I am using Nlogit 5 version to carry out estimations, but have > been following the Hensher et al. (2005) book to understand the process, > which might have used Nlogit 3 version. > > The book says (on pg. 398 and 399) that the Marginal effects as reported > are ?true marginal effects multiplied by 100?. So, we need to divide the > reported values by 100 to obtain the true marginal effect value. Now, I > presume that this is applicable to that particular version of the software > (i.e. Nlogit 3 or 4). > > > > However, I am using Nlogit 5 version, and it is not clear to me whether > the above convention is applicable to the marginal effects results given by > the Nlogit 5 version as well? > > I am asking this because the marginal effects obtained in my study are > already very small in magnitude. For example: they vary between 0.08 and > 1.06. If I divide such small values by 100, it will be exceptionally small. > > Is this not uncommon? > > Could you please advise on this? > > > > 2. The examples of marginal effects calculations given in the book, > Hensher et al. (2005), are for the MNL model specification. Do the syntax > remains the same if marginal effects are calculated for a mixed logit model? > > > > 3. Hensher et al. (2005) book has shown two different methods to > calculate marginal effects for continuous-level data and categorically > coded data. > > Details of attributes and levels adopted in my study are as follows: > > Alternatives: 2 > > Attributes: continuous (e.g. cost, time, etc.) ? 4 attributes with 3 > levels in each; categorical (i.e. qualitative attributes; e.g. type of a > facility, such as residential or commercial) ? 3 attributes with 2 levels > in each. > > Model type: Mixed logit model with heterogeneity in mean > > Marginal effect estimation method: ;Pwt (probability weighted average) > > > > Question : I have estimated the marginal effects of the continuous level > attributes, using the method given on page no. 397 (but applying changes to > the syntax that are required to estimate a mixed logit model). > > To estimate the marginal effects for the > categorically coded variables, the book shows the ?Simulation? method to be > used. > > > > However, when we use these two different methods separately to estimate > marginal effects (i.e. once for continuous variables and then for > categorical variables), two sets of coefficient values (i.e. parameter > estimates for the model) are obtained. But we need only one set of > coefficients for the estimated model. Which one of those two sets of > estimates are to be adopted for the final reporting? > > It will be a great help if you clarify these questions. I am happy to > provide more details, if needed. > > Kind regards > Ashu > > Ashu Kedia > PhD Student, > Department of Civil and Natural Resources Engineering, > University of Canterbury, Christchurch 8140 > New Zealand > ________________________________ > From: Ashu Kedia > Sent: Tuesday, December 10, 2019 12:57 PM > To: limdep at limdep.itls.usyd.edu.au > Subject: Request to help with estimating marginal effects > > Hi Professor William Greene, > I hope you are doing well. > > I am Ashu Kedia, a PhD student in transportation engineering at the > University of Canterbury, New Zealand. > > I am having a few doubts regarding marginal effects calculation. It will > be a great help if you could clarify my doubts. > > > 1. I am using Nlogit 5 version to carry out estimations, but have > been following the Hensher et al. (2005) book to understand the process, > which might have used Nlogit 3 version. > > The book says (on pg. 398 and 399) that the Marginal effects as reported > are ?true marginal effects multiplied by 100?. So, we need to divide the > reported values by 100 to obtain the true marginal effect value. Now, I > presume that this is applicable to that particular version of the software > (i.e. Nlogit 3 or 4). > > > > However, I am using Nlogit 5 version, and it is not clear to me whether > the above convention is applicable to the marginal effects results given by > the Nlogit 5 version as well? > > I am asking this because the marginal effects obtained in my study are > already very small in magnitude. For example: they vary between 0.08 and > 1.06. If I divide such small values by 100, it will be exceptionally small. > > Is this not uncommon? > > Could you please advise on this? > > > > 2. The examples of marginal effects calculations given in the book, > Hensher et al. (2005), are for the MNL model specification. Do the syntax > remains the same if marginal effects are calculated for a mixed logit model? > > > > 3. Hensher et al. (2005) book has shown two different methods to > calculate marginal effects for continuous-level data and categorically > coded data. > > Details of attributes and levels adopted in my study are as follows: > > Alternatives: 2 > > Attributes: continuous (e.g. cost, time, etc.) ? 4 attributes with 3 > levels in each; categorical (i.e. qualitative attributes; e.g. type of a > facility, such as residential or commercial) ? 3 attributes with 2 levels > in each. > > Model type: Mixed logit model with heterogeneity in mean > > Marginal effect estimation method: ;Pwt (probability weighted average) > > > > Question : I have estimated the marginal effects of the continuous level > attributes, using the method given on page no. 397 (but applying changes to > the syntax that are required to estimate a mixed logit model). > > To estimate the marginal effects for the > categorically coded variables, the book shows the ?Simulation? method to be > used. > > > > However, when we use these two different methods separately to estimate > marginal effects (i.e. once for continuous variables and then for > categorical variables), two sets of coefficient values (i.e. parameter > estimates for the model) are obtained. But we need only one set of > coefficients for the estimated model. Which one of those two sets of > estimates are to be adopted for the final reporting? > > It will be a great help if you clarify these questions. I am happy to > provide more details, if needed. > > Kind regards > Ashu > > > > _______________________________________________ > Limdep site list > Limdep at mailman.sydney.edu.au > http://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/7AcWClxwB5CV1N0zFG8m2u?domain=people.stern.nyu.edu Email: wgreene at stern.nyu.edu Ph. +1.646.596.3296 Editor in Chief: Journal of Productivity Analysis Editor in Chief: Foundations and Trends in Econometrics Associate Editor: Economics Letters Associate Editor: Journal of Business and Economic Statistics From dgiantsios at uom.gr Wed Dec 18 03:52:10 2019 From: dgiantsios at uom.gr (dgiantsios at uom.gr) Date: Tue, 17 Dec 2019 18:52:10 +0200 Subject: [Limdep Nlogit List] Bhargava/Sargan Test Message-ID: <9f1fac425ae97ce1fd073354c08047eb@uom.gr> Dear All Recently i run a dynamic random effects model with the two-step GMM estimation method because i include the lagged dependent variable in the regressors. LIMDEP/NLOGIT 5 gives the Bhargava/Sargan test but i cannot understand which H0 hypothesis this test searches. Is it preferable to obtain p-values for this statistic near zero or near one in order to have accurate estimates? Thanks n Advance From m.umair.k104 at gmail.com Wed Dec 18 16:11:41 2019 From: m.umair.k104 at gmail.com (Muhammad Umair Khan) Date: Wed, 18 Dec 2019 16:11:41 +1100 Subject: [Limdep Nlogit List] Simulation Scenario In-Reply-To: <5DF85BC8.2010207@sydney.edu.au> References: <5DF85BC8.2010207@sydney.edu.au> Message-ID: Thank you David for your response. Lastly, could you please comment on my interpretation of the output after I used the following command. ;simulate ;scenario:sidemi(meda,medb)=1 ;arc ;full ;pwt Attr Changed in | Change in Probability of Alternative ---------------------------------------------------------------------------- Choice MEDA | MEDA MEDB NONE x = SIDEMI | -.090 -.024 .202 Choice MEDB | MEDA MEDB NONE x = SIDEMI | -.026 -.064 .174 My interpretation: One unit increase in SIDEMI, that is from 0 (no side-effects) to 1 (mild side-effects) would decrease the probability of choosing Med A by 0.90 and Med B by 0.026 while the probability of choosing no medication (none) would increase by 0.202 and 0.174 for MedA and MedB respectively. Your comment will be highly appreciated. Thanks, Umair On Tue, Dec 17, 2019 at 3:38 PM David Hensher wrote: > I think this will result in 0 and 1 becoming 1 and 2 > > so instead use =1 > > > David Hensher > On 17/12/2019 3:33 PM, Muhammad Umair Khan via Limdep wrote: > > Dear Bill, > > > > Thank you very much for your reply. Can I also use the [+]1 scenario > > command in case of dummy-coded variables? For example, I have a > dummy-coded > > variable, mild side-effects, coded as 0 (no side-effects) and 1 (mild > > side-effects). Can I simulate a scenario "=[+]1 interpreting that there > > would be one level increase in the severity of mild side-effects, that > is, > > the severity would increase from to mild to, let's say, moderate? > > > > Thank you for your time. > > > > Umair > > > > On Thu, Dec 12, 2019 at 4:57 AM William Greene via Limdep< > > limdep at mailman.sydney.edu.au> wrote: > > > > > >> Umair. Yes, it will change a 1 to 3. It will also change a 2 to a 4. > >> Regards > >> Bill Greene > >> > >> On Tue, Dec 10, 2019 at 11:03 PM Muhammad Umair Khan via Limdep< > >> limdep at mailman.sydney.edu.au> wrote: > >> > >> > >>> Dear All, > >>> > >>> I am a newbie and just started using DCE and NLOGIT. I have three > >>> alternatives; Medication A, Medication B, No medication. One of my > >>> attributes is side-effects (SE) (0= no SE, 1= Mild SE, 2= Moderate SE, > 3= > >>> severe SE). I am trying to create a simulation scenario by checking the > >>> probability of taking medication if the severity of side-effects change > >>> from 1 to 3. I am using the following command: > >>> > >>> ;simulate > >>> ;scenario:sidemi(*)=[+]2 > >>> > >>> I am assuming that by using [+]2 NLOGIT would assume that I want to go > >>> > >> two > >> > >>> levels up from 1, that is, from 1 to 3 (mild to severe). Am I correct > >>> > >> here? > >> > >>> If no, what should be the best way? Furthermore the output I am getting > >>> > >> is, > >> > >>> for example, the probability of people taking medication would reduce > by > >>> 10%. I am interesting in finding out if this 10% is statistically > >>> significant? Any command through which we can get additional statistics > >>> such as p-value? > >>> > >>> Thank you! > >>> > >>> Umair > >>> _______________________________________________ > >>> Limdep site list > >>> Limdep at mailman.sydney.edu.au > >>> http://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/JSXECmOxDQtgWNrRfGzTQ6?domain=people.stern.nyu.edu > >> Email: wgreene at stern.nyu.edu > >> Ph. +1.646.596.3296 > >> Editor in Chief: Journal of Productivity Analysis > >> Editor in Chief: Foundations and Trends in Econometrics > >> Associate Editor: Economics Letters > >> Associate Editor: Journal of Business and Economic Statistics > >> _______________________________________________ > >> Limdep site list > >> Limdep at mailman.sydney.edu.au > >> http://limdep.itls.usyd.edu.au > >> > >> > >> > > _______________________________________________ > > Limdep site list > > Limdep at mailman.sydney.edu.au > > http://limdep.itls.usyd.edu.au > > > > > > > -- > DAVID HENSHER FASSA, PhD| Professor and Founding Director > Institute of Transport and Logistics Studies | The University of Sydney > Business School > > THE UNIVERSITY OF SYDNEY > Rm 201, Building H73| The University of Sydney | NSW | 2006 > Street Address: 378 Abercrombie St, Darlington NSW 2008 > T +61 2 9114 1871 | F +61 2 9114 1863 | M +61 418 433 057 > E David.Hensher at sydney.edu.au | W https://protect-au.mimecast.com/s/Sc_fCnxyErCPmNY6tJDy26?domain=sydney.edu.au |W > https://protect-au.mimecast.com/s/CG-JCoVzGQigv30KfVSAee?domain=sydney.edu.au > > Celebrating 25 years of ITLS: 1991-2016 https://protect-au.mimecast.com/s/dt8oCp8AJQtVA3BxsG07tL?domain=youtu.be > ERA Rank 5 (Transportation and Freight Services) > Co-Founder of the International Conference Series on Competition and > Ownership of Land Passenger Transport (The 'Thredbo' Series) > https://protect-au.mimecast.com/s/N3ZTCq7BKYtEXjo7fNSG70?domain=thredbo-conference-series.org > https://www.linkedin.com/company/28450714 > Second edition of Applied Choice Analysis now available at > https://protect-au.mimecast.com/s/xo21Cr8DLRtk2lGDtNASOq?domain=cambridge.org > Nlogit is the most popular software for choice modellers. See > https://protect-au.mimecast.com/s/PYoJCvl0PoCYAz3yHyem_y?domain=limdep.com > See Master of Transport beginning 2020: > * Master of Transport: > https://sydney.edu.au/courses/courses/pc/master-of-transport.html > * Graduate Certificate in Transport: > https://sydney.edu.au/courses/courses/pc/graduate-certificate-in-transport.html > * Graduate Diploma in Transport: > https://sydney.edu.au/courses/courses/pc/graduate-diploma-in-transport.html > > CRICOS 00026A > This email plus any attachments to it are confidential. Any unauthorised > use is strictly prohibited. If you receive this email in error, please > delete it and any attachments. > > Please think of our environment and only print this e-mail if necessary. > > > From mahsataslimi at gmail.com Fri Dec 20 03:49:49 2019 From: mahsataslimi at gmail.com (Mahsa Taslimi) Date: Thu, 19 Dec 2019 20:19:49 +0330 Subject: [Limdep Nlogit List] Nlogit ;pds Message-ID: Hi I am estimating RPL, LC, and LC-RPL models. I have a question about the observation and sample size. h have 2352 observations which are 784 in Nlogit. and the RpL got ; PDS for the fixed variable or the grouping. the estimation says I have 784 observations and the RPL or LC models got 98 groups and 2= pds. actually the WTP wil be 392 rows. what is this 392? is it the observations? or what? How does it work? i am learning it by myself. can anyone help me with this? Thanks Mahsa From wgreene at stern.nyu.edu Fri Dec 20 04:57:18 2019 From: wgreene at stern.nyu.edu (William Greene) Date: Thu, 19 Dec 2019 12:57:18 -0500 Subject: [Limdep Nlogit List] Nlogit ;pds In-Reply-To: References: Message-ID: Mahsa. For commands that have verb NLOGIT, the ;PDS is the number of choice tasks. One of my data sets has 12,800 rows of data. There are 400 individuals in the sample. Each individual faces 8 choice tasks. Each choice task involves 4 alternatives. So, the number of choice tasks in the data set is 12,800/4 = 3,200. For purposes of "panel data," the PDS for NLOGIT commands would be ;Pds=8. Note that ;PDS for other commands such as PROBIT would be Pds=32, but of course, this is wrong for multinomial choice data. /B. Greene On Thu, Dec 19, 2019 at 11:50 AM Mahsa Taslimi via Limdep < limdep at mailman.sydney.edu.au> wrote: > Hi > > I am estimating RPL, LC, and LC-RPL models. I have a question about the > observation and sample size. > h have 2352 observations which are 784 in Nlogit. and the RpL got ; PDS for > the fixed variable or the grouping. > the estimation says I have 784 observations and the RPL or LC models got 98 > groups and 2= pds. actually the WTP wil be 392 rows. what is this 392? is > it the observations? or what? > > How does it work? i am learning it by myself. can anyone help me with this? > > Thanks > Mahsa > _______________________________________________ > Limdep site list > Limdep at mailman.sydney.edu.au > http://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/zOaRC2xZYvCB7EzXtn0pNL?domain=people.stern.nyu.edu Email: wgreene at stern.nyu.edu Ph. +1.646.596.3296 Editor in Chief: Journal of Productivity Analysis Editor in Chief: Foundations and Trends in Econometrics Associate Editor: Economics Letters Associate Editor: Journal of Business and Economic Statistics From mahsataslimi at gmail.com Fri Dec 20 05:52:15 2019 From: mahsataslimi at gmail.com (Mahsa Taslimi) Date: Thu, 19 Dec 2019 22:22:15 +0330 Subject: [Limdep Nlogit List] Nlogit ;pds In-Reply-To: References: Message-ID: Thanks, Dr. Greene for replaying. I have two more questions left, I have 98 individuals and 8 choice cards and 3 alternatives in each choice cards and 2352 rows. all Data are the panel. I am using RPLOGIT, LCLOGIT, and LCRPLOGIT commands in the NLOGIT. based on your comment, I should use;pds=8. is it right? But in LCRPLOGIT the results of estimation says that the N=784, fixed number of observations is 2, the lcm model with panel 392 groups and the WTP are for 392 groups. how should I interpret 392? I don't know what is it? Thanks again Mahsa On Thu, Dec 19, 2019 at 9:28 PM William Greene via Limdep < limdep at mailman.sydney.edu.au> wrote: > Mahsa. For commands that have verb NLOGIT, the ;PDS is the number of > choice > tasks. One of my data sets has 12,800 rows of data. There are 400 > individuals in > the sample. Each individual faces 8 choice tasks. Each choice task > involves 4 > alternatives. So, the number of choice tasks in the data set is 12,800/4 = > 3,200. > For purposes of "panel data," the PDS for NLOGIT commands would be ;Pds=8. > Note that ;PDS for other commands such as PROBIT would be Pds=32, but of > course, > this is wrong for multinomial choice data. > /B. Greene > > On Thu, Dec 19, 2019 at 11:50 AM Mahsa Taslimi via Limdep < > limdep at mailman.sydney.edu.au> wrote: > > > Hi > > > > I am estimating RPL, LC, and LC-RPL models. I have a question about the > > observation and sample size. > > h have 2352 observations which are 784 in Nlogit. and the RpL got ; PDS > for > > the fixed variable or the grouping. > > the estimation says I have 784 observations and the RPL or LC models got > 98 > > groups and 2= pds. actually the WTP wil be 392 rows. what is this 392? is > > it the observations? or what? > > > > How does it work? i am learning it by myself. can anyone help me with > this? > > > > Thanks > > Mahsa > > _______________________________________________ > > Limdep site list > > Limdep at mailman.sydney.edu.au > > http://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/4PuhC71ZgLt8LQ76F83jZf?domain=people.stern.nyu.edu > Email: wgreene at stern.nyu.edu > Ph. +1.646.596.3296 > Editor in Chief: Journal of Productivity Analysis > Editor in Chief: Foundations and Trends in Econometrics > Associate Editor: Economics Letters > Associate Editor: Journal of Business and Economic Statistics > _______________________________________________ > Limdep site list > Limdep at mailman.sydney.edu.au > http://limdep.itls.usyd.edu.au > > From wgreene at stern.nyu.edu Fri Dec 20 06:02:31 2019 From: wgreene at stern.nyu.edu (William Greene) Date: Thu, 19 Dec 2019 14:02:31 -0500 Subject: [Limdep Nlogit List] Nlogit ;pds In-Reply-To: References: Message-ID: Mahsa. You can ignore the 392 (= 784/2) You should be using PDS=8 in all cases. The problem is that the program is reconciling the general use of PDS for by LIMDEP with the special counting rule for the NLOGIT multinomial choice commands. I'm assuming that you really do have 3 alternatives every time and that the choice set is correctly set for all the choice models. ./B. Greene On Thu, Dec 19, 2019 at 1:53 PM Mahsa Taslimi via Limdep < limdep at mailman.sydney.edu.au> wrote: > Thanks, Dr. Greene for replaying. > I have two more questions left, I have 98 individuals and 8 choice cards > and 3 alternatives in each choice cards and 2352 rows. all Data are the > panel. > > I am using RPLOGIT, LCLOGIT, and LCRPLOGIT commands in the NLOGIT. based on > your comment, I should use;pds=8. is it right? > But in LCRPLOGIT the results of estimation says that the N=784, fixed > number of observations is 2, the lcm model with panel 392 groups and the > WTP are for 392 groups. how should I interpret 392? I don't know what is > it? > > Thanks again > Mahsa > > > > On Thu, Dec 19, 2019 at 9:28 PM William Greene via Limdep < > limdep at mailman.sydney.edu.au> wrote: > > > Mahsa. For commands that have verb NLOGIT, the ;PDS is the number of > > choice > > tasks. One of my data sets has 12,800 rows of data. There are 400 > > individuals in > > the sample. Each individual faces 8 choice tasks. Each choice task > > involves 4 > > alternatives. So, the number of choice tasks in the data set is > 12,800/4 = > > 3,200. > > For purposes of "panel data," the PDS for NLOGIT commands would be > ;Pds=8. > > Note that ;PDS for other commands such as PROBIT would be Pds=32, but of > > course, > > this is wrong for multinomial choice data. > > /B. Greene > > > > On Thu, Dec 19, 2019 at 11:50 AM Mahsa Taslimi via Limdep < > > limdep at mailman.sydney.edu.au> wrote: > > > > > Hi > > > > > > I am estimating RPL, LC, and LC-RPL models. I have a question about the > > > observation and sample size. > > > h have 2352 observations which are 784 in Nlogit. and the RpL got ; PDS > > for > > > the fixed variable or the grouping. > > > the estimation says I have 784 observations and the RPL or LC models > got > > 98 > > > groups and 2= pds. actually the WTP wil be 392 rows. what is this 392? > is > > > it the observations? or what? > > > > > > How does it work? i am learning it by myself. can anyone help me with > > this? > > > > > > Thanks > > > Mahsa > > > _______________________________________________ > > > Limdep site list > > > Limdep at mailman.sydney.edu.au > > > http://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/oFWtCvl0PoCYv4GRhQ2fjo?domain=people.stern.nyu.edu > > Email: wgreene at stern.nyu.edu > > Ph. +1.646.596.3296 > > Editor in Chief: Journal of Productivity Analysis > > Editor in Chief: Foundations and Trends in Econometrics > > Associate Editor: Economics Letters > > Associate Editor: Journal of Business and Economic Statistics > > _______________________________________________ > > Limdep site list > > Limdep at mailman.sydney.edu.au > > http://limdep.itls.usyd.edu.au > > > > > _______________________________________________ > Limdep site list > Limdep at mailman.sydney.edu.au > http://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/oFWtCvl0PoCYv4GRhQ2fjo?domain=people.stern.nyu.edu Email: wgreene at stern.nyu.edu Ph. +1.646.596.3296 Editor in Chief: Journal of Productivity Analysis Editor in Chief: Foundations and Trends in Econometrics Associate Editor: Economics Letters Associate Editor: Journal of Business and Economic Statistics From mahsataslimi at gmail.com Fri Dec 20 06:10:41 2019 From: mahsataslimi at gmail.com (Mahsa Taslimi) Date: Thu, 19 Dec 2019 22:40:41 +0330 Subject: [Limdep Nlogit List] Nlogit ;pds In-Reply-To: References: Message-ID: Thank you. it is very helpful. Best regards Mahsa On Thu, Dec 19, 2019 at 10:33 PM William Greene via Limdep < limdep at mailman.sydney.edu.au> wrote: > Mahsa. You can ignore the 392 (= 784/2) You should be using PDS=8 > in all cases. The problem is that the program is reconciling the general > use of PDS for by LIMDEP with the special counting rule for the NLOGIT > multinomial choice commands. I'm assuming that you really do have 3 > alternatives every time and that the choice set is correctly set for all > the > choice models. > ./B. Greene > > On Thu, Dec 19, 2019 at 1:53 PM Mahsa Taslimi via Limdep < > limdep at mailman.sydney.edu.au> wrote: > > > Thanks, Dr. Greene for replaying. > > I have two more questions left, I have 98 individuals and 8 choice cards > > and 3 alternatives in each choice cards and 2352 rows. all Data are the > > panel. > > > > I am using RPLOGIT, LCLOGIT, and LCRPLOGIT commands in the NLOGIT. based > on > > your comment, I should use;pds=8. is it right? > > But in LCRPLOGIT the results of estimation says that the N=784, fixed > > number of observations is 2, the lcm model with panel 392 groups and the > > WTP are for 392 groups. how should I interpret 392? I don't know what is > > it? > > > > Thanks again > > Mahsa > > > > > > > > On Thu, Dec 19, 2019 at 9:28 PM William Greene via Limdep < > > limdep at mailman.sydney.edu.au> wrote: > > > > > Mahsa. For commands that have verb NLOGIT, the ;PDS is the number of > > > choice > > > tasks. One of my data sets has 12,800 rows of data. There are 400 > > > individuals in > > > the sample. Each individual faces 8 choice tasks. Each choice task > > > involves 4 > > > alternatives. So, the number of choice tasks in the data set is > > 12,800/4 = > > > 3,200. > > > For purposes of "panel data," the PDS for NLOGIT commands would be > > ;Pds=8. > > > Note that ;PDS for other commands such as PROBIT would be Pds=32, but > of > > > course, > > > this is wrong for multinomial choice data. > > > /B. Greene > > > > > > On Thu, Dec 19, 2019 at 11:50 AM Mahsa Taslimi via Limdep < > > > limdep at mailman.sydney.edu.au> wrote: > > > > > > > Hi > > > > > > > > I am estimating RPL, LC, and LC-RPL models. I have a question about > the > > > > observation and sample size. > > > > h have 2352 observations which are 784 in Nlogit. and the RpL got ; > PDS > > > for > > > > the fixed variable or the grouping. > > > > the estimation says I have 784 observations and the RPL or LC models > > got > > > 98 > > > > groups and 2= pds. actually the WTP wil be 392 rows. what is this > 392? > > is > > > > it the observations? or what? > > > > > > > > How does it work? i am learning it by myself. can anyone help me with > > > this? > > > > > > > > Thanks > > > > Mahsa > > > > _______________________________________________ > > > > Limdep site list > > > > Limdep at mailman.sydney.edu.au > > > > http://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/ifFSCVAGXPtWMz0DCGI6at?domain=people.stern.nyu.edu > > > Email: wgreene at stern.nyu.edu > > > Ph. +1.646.596.3296 > > > Editor in Chief: Journal of Productivity Analysis > > > Editor in Chief: Foundations and Trends in Econometrics > > > Associate Editor: Economics Letters > > > Associate Editor: Journal of Business and Economic Statistics > > > _______________________________________________ > > > Limdep site list > > > Limdep at mailman.sydney.edu.au > > > http://limdep.itls.usyd.edu.au > > > > > > > > _______________________________________________ > > Limdep site list > > Limdep at mailman.sydney.edu.au > > http://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/ifFSCVAGXPtWMz0DCGI6at?domain=people.stern.nyu.edu > Email: wgreene at stern.nyu.edu > Ph. +1.646.596.3296 > Editor in Chief: Journal of Productivity Analysis > Editor in Chief: Foundations and Trends in Econometrics > Associate Editor: Economics Letters > Associate Editor: Journal of Business and Economic Statistics > _______________________________________________ > Limdep site list > Limdep at mailman.sydney.edu.au > http://limdep.itls.usyd.edu.au > > From mahsataslimi at gmail.com Tue Dec 24 12:42:20 2019 From: mahsataslimi at gmail.com (Mahsa Taslimi) Date: Tue, 24 Dec 2019 05:12:20 +0330 Subject: [Limdep Nlogit List] Nlogit ;pds In-Reply-To: References: Message-ID: Hi again, I got a problem with Estimating WTP in LCRPL model. it is overestimated. I have 3 levels for the price: 400 (status quo), 1000 and 2000. The MWTP is above 2000. I don't know what to do, Thanks On Thu, Dec 19, 2019 at 10:40 PM Mahsa Taslimi wrote: > Thank you. it is very helpful. > > Best regards > Mahsa > > On Thu, Dec 19, 2019 at 10:33 PM William Greene via Limdep < > limdep at mailman.sydney.edu.au> wrote: > >> Mahsa. You can ignore the 392 (= 784/2) You should be using PDS=8 >> in all cases. The problem is that the program is reconciling the general >> use of PDS for by LIMDEP with the special counting rule for the NLOGIT >> multinomial choice commands. I'm assuming that you really do have 3 >> alternatives every time and that the choice set is correctly set for all >> the >> choice models. >> ./B. Greene >> >> On Thu, Dec 19, 2019 at 1:53 PM Mahsa Taslimi via Limdep < >> limdep at mailman.sydney.edu.au> wrote: >> >> > Thanks, Dr. Greene for replaying. >> > I have two more questions left, I have 98 individuals and 8 choice cards >> > and 3 alternatives in each choice cards and 2352 rows. all Data are the >> > panel. >> > >> > I am using RPLOGIT, LCLOGIT, and LCRPLOGIT commands in the NLOGIT. >> based on >> > your comment, I should use;pds=8. is it right? >> > But in LCRPLOGIT the results of estimation says that the N=784, fixed >> > number of observations is 2, the lcm model with panel 392 groups and >> the >> > WTP are for 392 groups. how should I interpret 392? I don't know what is >> > it? >> > >> > Thanks again >> > Mahsa >> > >> > >> > >> > On Thu, Dec 19, 2019 at 9:28 PM William Greene via Limdep < >> > limdep at mailman.sydney.edu.au> wrote: >> > >> > > Mahsa. For commands that have verb NLOGIT, the ;PDS is the number of >> > > choice >> > > tasks. One of my data sets has 12,800 rows of data. There are 400 >> > > individuals in >> > > the sample. Each individual faces 8 choice tasks. Each choice task >> > > involves 4 >> > > alternatives. So, the number of choice tasks in the data set is >> > 12,800/4 = >> > > 3,200. >> > > For purposes of "panel data," the PDS for NLOGIT commands would be >> > ;Pds=8. >> > > Note that ;PDS for other commands such as PROBIT would be Pds=32, but >> of >> > > course, >> > > this is wrong for multinomial choice data. >> > > /B. Greene >> > > >> > > On Thu, Dec 19, 2019 at 11:50 AM Mahsa Taslimi via Limdep < >> > > limdep at mailman.sydney.edu.au> wrote: >> > > >> > > > Hi >> > > > >> > > > I am estimating RPL, LC, and LC-RPL models. I have a question about >> the >> > > > observation and sample size. >> > > > h have 2352 observations which are 784 in Nlogit. and the RpL got ; >> PDS >> > > for >> > > > the fixed variable or the grouping. >> > > > the estimation says I have 784 observations and the RPL or LC models >> > got >> > > 98 >> > > > groups and 2= pds. actually the WTP wil be 392 rows. what is this >> 392? >> > is >> > > > it the observations? or what? >> > > > >> > > > How does it work? i am learning it by myself. can anyone help me >> with >> > > this? >> > > > >> > > > Thanks >> > > > Mahsa >> > > > _______________________________________________ >> > > > Limdep site list >> > > > Limdep at mailman.sydney.edu.au >> > > > http://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/QYeICgZowLHorAK5sN339Q?domain=people.stern.nyu.edu >> > > Email: wgreene at stern.nyu.edu >> > > Ph. +1.646.596.3296 >> > > Editor in Chief: Journal of Productivity Analysis >> > > Editor in Chief: Foundations and Trends in Econometrics >> > > Associate Editor: Economics Letters >> > > Associate Editor: Journal of Business and Economic Statistics >> > > _______________________________________________ >> > > Limdep site list >> > > Limdep at mailman.sydney.edu.au >> > > http://limdep.itls.usyd.edu.au >> > > >> > > >> > _______________________________________________ >> > Limdep site list >> > Limdep at mailman.sydney.edu.au >> > http://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/QYeICgZowLHorAK5sN339Q?domain=people.stern.nyu.edu >> Email: wgreene at stern.nyu.edu >> Ph. +1.646.596.3296 >> Editor in Chief: Journal of Productivity Analysis >> Editor in Chief: Foundations and Trends in Econometrics >> Associate Editor: Economics Letters >> Associate Editor: Journal of Business and Economic Statistics >> _______________________________________________ >> Limdep site list >> Limdep at mailman.sydney.edu.au >> http://limdep.itls.usyd.edu.au >> >> From aappiah at ualberta.ca Tue Dec 24 12:44:09 2019 From: aappiah at ualberta.ca (Alfred Appiah) Date: Mon, 23 Dec 2019 18:44:09 -0700 Subject: [Limdep Nlogit List] Unsubscribe Message-ID: Unsubscribe -- Alfred Appiah 587-778-3278 From mahsataslimi at gmail.com Tue Dec 24 14:51:58 2019 From: mahsataslimi at gmail.com (Mahsa Taslimi) Date: Tue, 24 Dec 2019 07:21:58 +0330 Subject: [Limdep Nlogit List] WTP Message-ID: Hi I got a problem with Estimating WTP in the LCRPL model. it is overestimated. I have 3 levels for the price: 400 (status quo), 1000 and 2000. The MWTP is above 2000. LCRPLogit ;lhs=choice ;choices=type1,type2,type3 ;model: u(type*)= a1*fn+a2*mhz+a3*egh+a4*si+a5*ej+a6*pr ;rpl ;halton ;draws=500 ;fcn=a5(n) ;pds=8 ;lcm ;pts=2 ;Par ;wtp = a1/a6, a2/a6, a3/a6, a4/a6, a5/a6 ;Parameters ;Effects : fn(*)/pr(*), mhz(*)/pr(*), egh(*)/pr(*), ej(*)/pr(*), si(*)/pr(*) $ MATRIX ;list ;1/79 *wtp_i'1 $ MATRIX ;list ;1/79 *classp_i' 1 $ can you please help me with this? From fengxiaz at gmail.com Tue Dec 24 14:52:43 2019 From: fengxiaz at gmail.com (fengxia zhu) Date: Mon, 23 Dec 2019 22:52:43 -0500 Subject: [Limdep Nlogit List] Saved Matrices for Average Partial Effects Coefficients Message-ID: Hi, I'm using Fractional Response Model with the "Partial" command. I would like to use the Average Partial Effects coefficients ( the coefficients and the variance of the coefficients) for subsequent bootstrapping procedures but couldn't find them under "projects-->matrices-->B/VarB". It appears that only the coefficients for conditional mean and group mean are stored. Is there any way to save/find the B and VarB for the APE coefficients? Any suggestions are appreciated! Sandy Cleveland State University