mode choice experiment

[Hamad Almujibah]

I am a new in using the Stata software and dealing with a mode choice experiment having three different alternatives of transport modes with three different attributes of cost, frequency, and time in order to select the most suitable mode based on the given variables. I use the following function in the command:

asclogit Chosen Cost Freq Time, case(Observation) alternatives(Alternative)

and I get a note that is saying: variable Freq has 152 cases that are not alternative-specific: there is no within-case variability

It will be highly appreciated if can someone helps with this issue.

Hamad

[Marcos Almeida]

Welcome to the Stata Forum / Statalist.

With regards to the note, it seems Stata is telling you that there is an issue. Being this true, i.e., if there is no alternative within case, you may wish to think about reconsidering the model selection.

By the way, 152 cases are a good fraction of the sample, or just a tiny portion?

[Hamad Almujibah]

Iteration 0: log likelihood = -3604.6696
Iteration 1: log likelihood = -3542.0429
Iteration 2: log likelihood = -3537.2062
Iteration 3: log likelihood = -3537.2013
Iteration 4: log likelihood = -3537.2013

Alternative-specific conditional logit Number of obs = 12,312
Case variable: Observation Number of cases = 4104

Alternative variable: Alternative Alts per case: min = 3
avg = 3.0
max = 3

Wald chi2(3) = 26.74
Log likelihood = -3537.2013 Prob > chi2 = 0.0000

------------------------------------------------------------------------------
Chosen | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
Alternative |
Cost | -.0000184 .0005372 -0.03 0.973 -.0010713 .0010345
Freq | -.0007146 .0014547 -0.49 0.623 -.0035658 .0021366
Time | -.370864 .0718392 -5.16 0.000 -.5116662 -.2300617
-------------+----------------------------------------------------------------
A | (base alternative)
-------------+----------------------------------------------------------------
B |
_cons | .6565061 .3197663 2.05 0.040 .0297756 1.283237
-------------+----------------------------------------------------------------
C |
_cons | 1.247181 .1488171 8.38 0.000 .9555047 1.538857
------------------------------------------------------------------------------

[Marcos Almeida]

It seems there is a large sample and such special cases may be taken as a small fraction. There are similar examples in the Web, like this one.

[Hamad Almujibah]

I have tried but it seems there is something unreadable with the frequency because the p>z of other cost and time are equal to 0.000 as seen

. asclogit Chosen Cost Freq Time, case(Observation) alternatives(Alternative) nocons
note: variable Freq has 152 cases that are not alternative-specific: there is no within-case variability

Iteration 0: log likelihood = -3889.6304
Iteration 1: log likelihood = -3743.9506
Iteration 2: log likelihood = -3739.0069
Iteration 3: log likelihood = -3739.0011
Iteration 4: log likelihood = -3739.0011

Alternative-specific conditional logit Number of obs = 12,312
Case variable: Observation Number of cases = 4104

Alternative variable: Alternative Alts per case: min = 3
avg = 3.0
max = 3

Wald chi2(3) = 1245.58
Log likelihood = -3739.0011 Prob > chi2 = 0.0000

------------------------------------------------------------------------------
Chosen | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
Alternative |
Cost | -.0050906 .0004607 -11.05 0.000 -.0059936 -.0041877
Freq | .0013877 .0014398 0.96 0.335 -.0014343 .0042098
Time | -.5553529 .0214295 -25.92 0.000 -.597354 -.5133518
------------------------------------------------------------------------------

[Hamad Almujibah]

As I have two experiments with the same strategy but different alternatives but for this one there is not note as shown in the one above
asclogit Chosen Cost Freq Time, case(Observation) alternatives(Alternative) nocons

Iteration 0: log likelihood = -5589.4314
Iteration 1: log likelihood = -5585.8716
Iteration 2: log likelihood = -5585.8681
Iteration 3: log likelihood = -5585.8681

Alternative-specific conditional logit Number of obs = 16,412
Case variable: Observation Number of cases = 4103

Alternative variable: Alternative Alts per case: min = 4
avg = 4.0
max = 4

Wald chi2(3) = 194.03
Log likelihood = -5585.8681 Prob > chi2 = 0.0000

------------------------------------------------------------------------------
Chosen | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
Alternative |
Cost | -.0030127 .0003406 -8.84 0.000 -.0036803 -.0023451
Freq | .0003707 .0006293 0.59 0.556 -.0008627 .0016041
Time | -.437172 .0345546 -12.65 0.000 -.5048977 -.3694462
------------------------------------------------------------------------------

Your help is highly appreciated and thanks a lot