Priors from estimating the DCE data

[rtang]

Dear admin and all peer-users of Ngene,

I am a little confused by the prior values required by Ngene. Hope someone may offer clues.

With a DCE dataset, I normally do not estimate a parameter of each variable directly, particularly when the levels of an attribute is not linear. Instead, I get coefficients of the levels. For example, if my model is:
U(alt1) = b1*A[0, 1, 2, 3] + b2*B[0,1]
U(alt2) = b1*A+b2*B

If A is “colour” and has four levels (red, black, white, green), B is “car” with ‘Honda’ and ‘Benz’, then based on the DCE dataset, I will have a clogit output like:
Choice = cl1*red+ cl2*black+ cl3*white+c1*Honda
rather than: Choice = cl*colour+ c*car.

So, I won’t be able to have ‘b1’ and ‘b2’. How can I set priors in Ngene?

[Andrew Collins]

This is covered in detail in the manual. See section 7.2.8, Designs with dummy and effects coded attributes.

Andrew