[Limdep Nlogit List] Base model in simulation

[ho quoc Chinh]

Dear Limdep users,
 
I used the CLOGIT.DAT data to estimate the MNL and NL models and then compared the two results with each other in terms of scenario analysis. However, there is an inconsistency between base model in Simulation command and Constant only model in NL model. The code and results are shown as follows:
 
Code In Simulation Command
NLOGIT; Lhs= MODE;
 Choices=Air, Train, Bus, Car;
 RU2;
 Tree=Flight(Air),Ground(Train, Bus, Car);
 Model:    
   U(Air) = AA+BC*GC+BT*TTME+Bh*HINC/ 
   U(Train) = AT+BC*GC+BT*TTME/ 
   U(Bus)  = AB+BC*GC+BT*TTME/
   U(Car)  =    BC*GC+BT*TTME$
NLOGIT; Lhs= MODE;
 Choices=Air, Train, Bus, Car;
 RU2;
 Tree=Flight(Air),Ground(Train, Bus, Car);
 Model: U(Air) = AA+BC*GC+BT*TTME+Bh*HINC/ 
   U(Train) = AT+BC*GC+BT*TTME/ 
   U(Bus)  = AB+BC*GC+BT*TTME/
   U(Car)  =    BC*GC+BT*TTME;
Simulation;
Scenario:gc(*)=[*]3.0$
 
Results:








Choice
Base


%share
Number

air
21.86 
46 

train
30.92 
65 

bus
18.18 
38 

car
29.05 
61 

Total
100.00 
210

In Nested Logit constant only model
NLOGIT; Lhs= MODE;
 Choices=Air, Train, Bus, Car;
 RU2;
 Prob=predprob;
 Tree=Flight(Air),Ground(Train, Bus, Car);
 Rhs=one$
Results of predprob variable
0.27619 
0.3
0.142857
0.280952

Comparing the two results as shown above, we can see the far diffeences between the two constant only models (one is calculated directly by NL and the other by Simulation)
 
Where is the error in my code?
 
Furthermore, I have already compared the results of MNL models and they are consistent.
 
Any suggestion is hightly appreciated.

Best regards,
Chinh.
Department of Civil Engineering, Nagoya University.


      ____________________________________________________________
Thị trường chứng khoán Việt Nam hot tới mức nào?
Khám phá tại Yahoo! Hỏi & Đáp
http://vn.answers.yahoo.com