[Limdep Nlogit List] Chuyển tiếp: Problem with scenario analysis

[ho quoc Chinh]

Dear Lipdepers,
 
There is some error occuring in the last email. I forward it again for the sake of convenience.

Best regards,
HO Quoc Chinh
-------------------------------------------------------------------------
In Japan
Address: Nutrend Lab - Department of Civil Engineering, Nagoya University.
C1-3(651) Furo-cho, Chikusa-ku, Nagoya. 464-8603 Japan
Tel: 080-3620-8550

In Vietnam
Address: Bridge and Highway Engineering, Faculty of Civil Engineering, HCMUT.
268 Ly Thuong Kiet Str. Ward 15 - District 11 - Ho Chi Minh City - Vietnam
Tel: 0903-154-493

--- Ngày Thứ 4, 16/12/09, ho quoc Chinh <hoqchinh at yahoo.com.vn> đã viết:


Từ: ho quoc Chinh <hoqchinh at yahoo.com.vn>
Chủ đề: Problem with scenario analysis
Đến: limdep at limdep.itls.usyd.edu.au
Ngày: Thứ Tư, 16 tháng 12, 2009, 8:12







Dear Limdep users,
 
I encountered an unexpected problem with scenario analysis in Nested logit model using Nlogit version 3.0. That is the results of simulation change according to the order of nests (branches) in "tree" definition. Here is my command syntax and the corresponding results:
 
Model 1: 
 
NLOGIT
;Lhs= CHOICE
;Choices=None,b1,b2,mc1,b1mc1,b2mc1,mc2,b1mc2,b2mc2,car
;Tree=Poor(None,b1,b2,mc1),Rich(mc2,b1mc2,car),Mid(b1mc1,b2mc1,b2mc2)
;ivset:(Poor)=[1.0]
;Tlg=0.00001
;crosstable
;maxit=150
;start=logit
;Model:
U(none)=..../
U(b1)=....../ 
....
....
....
U(car)=......$

NLOGIT
;Lhs= CHOICE
?Similar as above
;Model:
U(none)=..../
U(b1)=....../ 
....
....
....
U(car)=......;
Simulation
;Scenario:INC(b1,b2,mc1,b1mc1,b2mc1,mc2,b1mc2,b2mc2,car)=[*]2.0/
UC_INC(b1,b2,mc1,b1mc1,b2mc1,mc2,b1mc2,b2mc2,car)=[*]0.5$
 
Model 1 simulation result:
 












Choice
Base
Scenario
Scenario-Share

 
%Share 
Number
%Share 
Number
ChgShare
ChgNumber

NONE
1.555
309
0.698
139
-0.86%
-170

B1
4.381
871
1.853
369
-2.53%
-502

B2
5.487
1092
1.983
394
-3.51%
-698

MC1
2.934
584
1.706
339
-1.23%
-245

B1MC1
7.372
1467
4.008
797
-3.37%
-670

B2MC1
7.161
1425
3.472
691
-3.69%
-734

MC2
21.39
4255
28.556
5681
7.17%
1426

B1MC2
44.214
8796
49.143
9777
4.93%
981

B2MC2
1.327
264
0.86
171
-0.47%
-93

CAR
4.179
831
7.72
1536
3.54%
705

Total
100
19894
100
19894
0.00%
0
 
Then I just change the order of Rich and Mid branches with each other in tree definition while keeping all others exactly same as model 1 
(i.e., ;Tree=Poor(None,b1,b2,mc1),Mid(b1mc1,b2mc1,b2mc2),Rich(mc2,b1mc2,car))
, I got a diffrent result
 
Model 2 simulation result:
 












Choice
Base
Scenario
Scenario-Share

 
%Share 
Number
%Share 
Number
ChgShare
ChgNumber

NONE
1.212
241
0.574
114
-0.64%
-127

B1
2.306
459
0.99
197
-1.32%
-262

B2
1.357
270
0.507
101
-0.85%
-169

MC1
2.148
427
1.608
320
-0.54%
-107

B1MC1
32.16
6398
26.578
5287
-5.58%
-1111

B2MC1
33.61
6686
23.528
4681
-10.08%
-2005

MC2
2.212
440
3.291
655
1.08%
215

B1MC2
2.198
437
2.619
521
0.42%
84

B2MC2
22.588
4494
39.911
7940
17.32%
3446

CAR
0.21
42
0.393
78
0.18%
36

Total
100
19894
100
19894
0.00%
0

Is there any one experiencing this problem? Any suggestion is highly appreciated.

Best regards,
HO Quoc Chinh
------------------------------------------------------------------------------------------------------------
In Japan
Address: Nutrend Lab - Department of Civil Engineering, Nagoya University.
C1-3(651) Furo-cho, Chikusa-ku, Nagoya. 464-8603 Japan
Tel: 080-3620-8550

In Vietnam
Address: Bridge and Highway Engineering, Faculty of Civil Engineering, HCMUT.
268 Ly Thuong Kiet Str. Ward 15 - District 11 - Ho Chi Minh City - Vietnam
Tel: 0903-154-493


Nước nào thường cho nhiều học bổng du học? Khám phá tại Yahoo! Hỏi & Đáp


      Truy cập Mail nhanh hơn! Yahoo! khuyến khích bạn nâng cấp trình duyệt lên Internet Explorer 8 mới, tối ưu hóa cho Yahoo!. Tải tại đây! 
http://downloads.yahoo.com/vn/internetexplorer/