[Limdep Nlogit List] Fwd: SP to RP scale parameter in combined mixed logit
Zeinab Yahyazadeh Jasour
zeinabj at udel.edu
Wed Jul 20 03:46:17 AEST 2016
Dear limdep user,
I have sent this question last week but I did not get any response back. I
really appreciate if anyone gives me any suggestion or recommends any
material that can help me find a solution for my problem? Moreover, I found
out that there is another model named GMXL to combine SP/RP, should I try
that?
---------- Forwarded message ----------
From: Zeinab Yahyazadeh Jasour <zeinabj at udel.edu>
Date: Thu, Jul 14, 2016 at 11:58 AM
Subject: SP to RP scale parameter in combined mixed logit
To: limdep at limdep.itls.usyd.edu.au
Dear limdep users,
I'm having trouble running a combined RP/SP mixed logit model. I cannot
find the SP to RP scale parameter. I tried checking the NLOGIT manuals and
"Applied Choice Analysis, 2nd edition”. On page 852 this book, the concept
of the SP scale parameter is discussed. It is stated there that you can
find the SP scale parameter based on standard deviation of ASC of
alternative, but this information is not in the output of our model.
The data I am analyzing comes from a discrete choice, choice experiment
survey. I have about 313 survey respondents that were each asked a yes or
no choice experiment question 3 times, one RP and two SP. My universal
choice set consists of 7 alternatives (one choice and a none alternative
for revealed preference, and four choices and a none for stated preference)
and each question consists of two alternatives (a choice and a none
alternative). Thus, I have about 1878 rows of data. This is my current
code.
NLOGIT
;Lhs=STRAP,nij,alt
;Choices=rpyes,rpno,spyes,spgyes,spryes,splyes, spno
;ecm=(rpno,spno),(rpyes,spyes),(spgyes),(spryes),(splyes)
;par
;halton;pts=150
;pds=3
;rpl
;model:
U(rpyes)=R1 +WI*windinc+chi*CHILD+RA*RACE/
U(rpno)=0/
U(spyes)=NOI1 +WI*windinc+chi*CHILD+RA*RACE/
U(spgyes)= G1 +WI*windinc+chi*CHILD+RA*RACE/
U(spRyes)= PR1 +WI*windinc+chi*CHILD+RA*RACE/
U(splyes)= LL1 +WI*windinc+chi*CHILD+RA*RACE/
U(spno)=0
;fcn=wi(n)$
And the output is as follow:
Random Parms/Error Comps. Logit Model
Dependent variable STRAP
Log likelihood function -7568.75368
Estimation based on N = 508, K = 14
Inf.Cr.AIC = 15165.5 AIC/N = 29.853
---------------------------------------
Log likelihood R-sqrd R2Adj
No coefficients -988.5224 ************
Constants only can be computed directly
Use NLOGIT ;...;RHS=ONE$
At start values -7887.7841 .0404 .0133
Note: R-sqrd = 1 - logL/Logl(constants)
Warning: Model does not contain a full
set of ASCs. R-sqrd is problematic. Use
model setup with ;RHS=one to get LogL0.
---------------------------------------
Response data are given as ind. choices
Replications for simulated probs. = 150
Used Halton sequences in simulations.
RPL model with panel has 313 groups
Fixed number of obsrvs./group= 3
BHHH estimator used for asymp. variance
Number of obs.= 939, skipped 431 obs
--------+--------------------------------------------------------------------
| Standard Prob. 95% Confidence
STRAP| Coefficient Error z |z|>Z* Interval
--------+--------------------------------------------------------------------
|Random parameters in utility
functions..............................
WI| -11.9917 .7322D+07 .00 1.0000 *********** ***********
|Nonrandom parameters in utility
functions...........................
R1| 11.4509 .7322D+07 .00 1.0000 *********** ***********
CHI| .15277 .40867 .37 .7085 -.64820 .95375
RA| -.49102 .55598 -.88 .3771 -1.58073 .59868
NOI1| 8.68247 .7322D+07 .00 1.0000 *********** ***********
G1| 3251.0 .1218D+14 .00 1.0000 -.23868D+14 .23868D+14
PR1| 27625.5 .9375D+13 .00 1.0000 *********** ***********
LL1| 166.500 .1128D+14 .00 1.0000 *********** ***********
|Distns. of RPs. Std.Devs or limits of
triangular....................
NsWI| .02295 23.31667 .00 .9992 -45.67688 45.72278
|Standard deviations of latent random
effects........................
SigmaE01| .02779 20.41008 .00 .9989 -39.97523 40.03081
SigmaE02| .03581 22.56064 .00 .9987 -44.18224 44.25387
SigmaE03| 0.0 .7832D+15 .00 1.0000 -.15351D+16 .15351D+16
SigmaE04| 0.0 .2999D+15 .00 1.0000 -.58777D+15 .58777D+15
SigmaE05| 0.0 .3843D+15 .00 1.0000 -.75323D+15 .75323D+15
--------+--------------------------------------------------------------------
nnnnn.D-xx or D+xx => multiply by 10 to -xx or +xx.
***, **, * ==> Significance at 1%, 5%, 10% level.
Model was estimated on Jul 14, 2016 at 10:17:19 AM
-----------------------------------------------------------------------------
I hope you can clarify this.
Thank you very much in advance!
More information about the Limdep
mailing list