From Lixian.Qian at xjtlu.edu.cn Tue May 7 17:51:26 2019 From: Lixian.Qian at xjtlu.edu.cn (Lixian Qian) Date: Tue, 7 May 2019 07:51:26 +0000 Subject: [Limdep Nlogit List] Negative standard deviation of error component Message-ID: <0f163d8d1dc545218d96b71d127a0b8b@xjtlu.edu.cn> Hello, We are running an error component logit model with four alternatives. SAMPLE ; All $ ECLOGIT; Lhs = Choice ; Choices=CB,BL,CL,CS ;Model: U(CB) =Price*Capital + TripCost*TripCost+subsidy*subsidy + licen*licen + FCSERVICE*CSERVICE + FCTIME*TSERVICE + home*home + renew*recost/ U(BL) =Price*Capital + TripCost*TripCost+subsidy*subsidy + licen*licen + BSERVICE*CSERVICE + BLTIME*TSERVICE + home*home + BL/ U(CL) =RentCost*Cost + licen*licen+FCSERVICE*CSERVICE + FCTIME*TSERVICE + home*home + CL/ U(CS) =ShareCost*TripCost + licen*licen + CSERVICE*CSERVICE + checka*checka+CS ; CheckData ; PTS=500 ; ECM=(CB),(BL),(CL),(CS) ; PDS=4 ; maxit=500 ; Output = IC$ To our surprise, the results show that one error component has negative (and significant) standard deviation. Can anyone please advise on the issue? Thank you. Best, Lixian From wgreene at stern.nyu.edu Tue May 7 22:10:37 2019 From: wgreene at stern.nyu.edu (William Greene) Date: Tue, 7 May 2019 08:10:37 -0400 Subject: [Limdep Nlogit List] Negative standard deviation of error component In-Reply-To: <0f163d8d1dc545218d96b71d127a0b8b@xjtlu.edu.cn> References: <0f163d8d1dc545218d96b71d127a0b8b@xjtlu.edu.cn> Message-ID: Lixian. In a 4 choice model, you can only have 3 independent error components. You should drop one of them. /B. Greene On Tue, May 7, 2019 at 4:06 AM Lixian Qian wrote: > Hello, > > We are running an error component logit model with four alternatives. > > SAMPLE ; All $ > ECLOGIT; Lhs = Choice ; Choices=CB,BL,CL,CS > ;Model: > U(CB) =Price*Capital + TripCost*TripCost+subsidy*subsidy + licen*licen > + FCSERVICE*CSERVICE + FCTIME*TSERVICE + home*home + renew*recost/ > U(BL) =Price*Capital + TripCost*TripCost+subsidy*subsidy + licen*licen > + BSERVICE*CSERVICE + BLTIME*TSERVICE + home*home + BL/ > U(CL) =RentCost*Cost + licen*licen+FCSERVICE*CSERVICE + > FCTIME*TSERVICE + home*home + CL/ > U(CS) =ShareCost*TripCost + licen*licen + CSERVICE*CSERVICE + > checka*checka+CS > ; CheckData > ; PTS=500 > ; ECM=(CB),(BL),(CL),(CS) > ; PDS=4 > ; maxit=500 > ; Output = IC$ > > To our surprise, the results show that one error component has negative > (and significant) standard deviation. > > Can anyone please advise on the issue? Thank you. > > > Best, > Lixian > > > _______________________________________________ > Limdep site list > Limdep at mailman.sydney.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: https://protect-au.mimecast.com/s/elrPCjZrzqHNDKkpFWXznm?domain=people.stern.nyu.edu Email: wgreene at stern.nyu.edu Ph. +1.212.998.0876 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 Associate Editor: Journal of Choice Modeling From richard.turner at imarketresearch.com Wed May 15 00:10:19 2019 From: richard.turner at imarketresearch.com (Richard Turner) Date: Tue, 14 May 2019 10:10:19 -0400 Subject: [Limdep Nlogit List] How to interpret significance in DCMs? Message-ID: Greetings, How should the analyst interpret changes in significant variables simply due to changing the base alternative in a discrete choice model?. In discrete choice models, significance of a variable seems to be influenced not only by the base level of the attribute, but also the base alternative. How can the analyst still be confident in the stability of the results? Regards, Richard From avassilopoulos.aua at gmail.com Thu May 16 01:10:07 2019 From: avassilopoulos.aua at gmail.com (Achilleas' gmail) Date: Wed, 15 May 2019 18:10:07 +0300 Subject: [Limdep Nlogit List] How to interpret significance in DCMs? In-Reply-To: References: Message-ID: <002401d50b30$493714e0$dba53ea0$@gmail.com> Hi Richard, Assuming that you ran a conditional logit, your results (in terms of predicted choice probabilities) will not change. The differences you observe in the magnitude and the statistical significance of the estimates when you alter the base alternative and the base level of an attribute are due to the change in the interpretation of coefficients (so in essence you are comparing different effects). You can use the ;WALD command to estimate the coefficients you would get under a different base alternative. Best, Achilleas -----Original Message----- From: Limdep On Behalf Of Richard Turner Sent: Tuesday, May 14, 2019 17:10 To: limdep at mailman.sydney.edu.au Subject: [Limdep Nlogit List] How to interpret significance in DCMs? Greetings, How should the analyst interpret changes in significant variables simply due to changing the base alternative in a discrete choice model?. In discrete choice models, significance of a variable seems to be influenced not only by the base level of the attribute, but also the base alternative. How can the analyst still be confident in the stability of the results? Regards, Richard _______________________________________________ Limdep site list Limdep at mailman.sydney.edu.au http://limdep.itls.usyd.edu.au