[Discrete Choice Models] Error when running cmxtmixlogit command

[Gabin Morillon]

Dear Statalist,

I'm using discrete choice modelling with the -cmxtmixlogit- provided by the software. The design of the experiment is as follow:
- 19 scenarios (choices)
- 2 alternatives without optout
- 7 attributes with 4 to 6 levels
Then, 38 lines by individual.

Here is an illustration for the first ID:

    id    group    alt    valu    
                    
1.    1    1    0    1    
2.    1    1    1    0    
3.    1    2    0    1    
4.    1    2    1    0    
5.    1    3    0    0    
6.    1    3    1    1    
7.    1    4    0    0    
8.    1    4    1    1    
9.    1    5    0    1    
10.    1    5    1    0    
11.    1    6    0    0    
12.    1    6    1    1    
13.    1    7    0    1    
14.    1    7    1    0    
15.    1    8    0    1    
16.    1    8    1    0    
17.    1    9    0    0    
18.    1    9    1    1    
19.    1    10    0    0    
20.    1    10    1    1    
21.    1    11    0    1    
22.    1    11    1    0    
23.    1    12    0    1    
24.    1    12    1    0    
25.    1    13    0    0    
26.    1    13    1    1    
27.    1    14    0    0    
28.    1    14    1    1    
29.    1    15    0    1    
30.    1    15    1    0    
31.    1    16    0    0    
32.    1    16    1    1    
33.    1    17    0    1    
34.    1    17    1    0    
35.    1    18    0    1    
36.    1    18    1    0    
37.    1    19    0    1    
38.    1    19    1    0

I cannot use the -mixlogit- command because it only accepts the Halton as integration method that does not support than 20 coefficients. So, I decided to use the -cmxtmixlogit- command that allows the random method to alleviate the large numbers of coefficients. Here is my code:

cmxtmixlogit valu,                   ///
    random(ib1.PF#c.du           ///
            ib1.RL#c.du                ///
            ib1.SF#c.du                ///
            ib1.PA#c.du                ///
            ib1.MH#c.du               ///
            ib1.VT#c.du                ///
             c.du, normal)             ///
                intmethod(random) tech() intpoints() intseed() favor()

Unfortunately, I got this error message:

Fitting fixed parameter model:

Fitting full model:
_ASMIX_lgprime():  3301    subscript invalid
_ASMIX_setup():     -    function returned    error
<istmt>:     -    function returned    error
r(3301);

Does anyone know something about it? I didn't found any cues. It seems that it is linked to the technical capacities of Stata.

Many thanks,

GM

[Joerg Luedicke (StataCorp)]

It looks as though your Stata is not up to date. Please open a fresh Stata instance and enter

update query

then follow the instructions on the screen. Once updated, try to run your code again. The behavior you observe is likely due to a problem that has since been fixed. In Stata, enter

help whatsnew17

and look for item 4 under update 08mar2023.

[Gabin Morillon]

It looks as though your Stata is not up to date. Please open a fresh Stata instance and enter

update query

then follow the instructions on the screen. Once updated, try to run your code again. The behavior you observe is likely due to a problem that has since been fixed. In Stata, enter

help whatsnew17

and look for item 4 under update 08mar2023.

Hi Joerg,

Thanks a lot, it perfectly works! I never looked at the Stata updates... Hope this post will help other Stata users.

Best,

Gabin