choice-metrics.com

Hi,

I notice Michel you said you don't use Nlogit, can I ask what software you use to analyze DCEs?
I am exploring open-access alternatives to Nlogit, (that are not overly complicated).

Thanks,

Edel

Nlogit is a commercial product that uses syntax similar to Ngene and is likely the easiest to use. It also comes with an extensive manual.

There are two main alternatives, both freely available:
1. Apollo, as a package in R
2. Biogeme, as a package in Python

Both Apollo and Biogeme do more or less the same thing and it is mainly a matter of whether you are more familiar with R or with Python. Both have example scripts to estimate models, so you can copy & paste most of the code without a real need for learning R or Python (although it would be useful if you want to estimate more flexible model types).

Michiel

Hi Michiel,

Thank you for your answer to my previous question.

I am running a pilot with no priors, 2 alternatives, 6 factors (4 with 3 levels, 2 with 2 levels). I keep on getting this error when I run the syntax, Error: The model has constants specified for all alternatives.
Can you identify where I am going wrong?

Design
;alts=intensive, standard
;rows=72
;eff= (mnl,d)
;block =6
;model:


U(intensive) = reg+
cvev[-0.00001] *CVEV[11,15,22]+
fall[-0.00001] *FALL[6,8,12]+
sae[-0.00001] *SAE[1,5,26]+
dizz[-0.00001] *DIZZ[1,8,43]+
dh.dummy[0.00001]* DH[1,0]+
supp.dummy[0.00001]* SUPP[1,0] /


U(standard) = reg[0] +
cvev * CVEV+
fall*FALL+
sae*SAE+
dizz*DIZZ+
dh* DH+
supp* SUPP


$

As the error message says, you added constants for all alternatives, which makes the model overspecified. In choice models only differences in utility matter, therefore you can only add a constant to either alternative 'intensive' or alternative 'standard'.

For any further Ngene specific questions, please create a separate post in the Ngene forum.

Michiel

Whoops that was an easy fix.
Thank you very much.
I will post on the Ngene board in future.