choice-metrics.com

Hi everyone
I am new to ngene. I want to design a DCE with two alternatives in which, the level of the first attribute is same for both alternatives in every choice task (level overlap).
I have read the manual, and used both cond and require properties .See the following code:

;require:
A.att1 = B.att1

Or
;cond:
if(A.att1=1,B.att1 =1),
if(A.att1=2,B.att1 =2),
if(A.att1=3,B.att1 =3)

None of these seems to work. It either doesn't give me the overlap design, or it gives the message "No valid design has been found." Could you think of any other suggestion?
Also, it takes more than 20 hours to run. Is it normal? Is there any option that I can shorten the time?

Thank you in advance

The easiest way to link attribute levels of multiple alternatives is using syntax like:

U(alt1) = b1*x1[1,2,3] ... /
U(alt2) = b2*x1[x1] ... /
U(alt3) = b3

This ensures that x1 has the same levels across alternatives 1 and 2. This is explained in the Ngene manual page 174.

Note that you can only estimate this model if there is a third alternative in which x1 does not appear or where it does not have the same value, as otherwise this attribute is constant across all alternatives and cannot be estimated. This is likely the reason why Ngene is not able to find a design. In order to resolve this, you can add a "no choice" alternative (which only has a constant). If you do not want to do this, you can also choose to use interaction terms, namely

U(alt1) = b1*x1[1,2,3]*x2[1,2,3] + ... /
U(alt2) = b1*x1[x1]*x2 + ... /

Michiel

Thank you very much for your help Michiel.
Will use this method.
I have an opt-out option but x1 has the same value for that one as well.
Because of the conceptual nature of our DCE, x1 has the same levels across all 3 alternatives. My main goal is to estimate the interaction between this attribute and other attributes.

Just to clarify, an opt out alternative does not have any attributes, so when you say that x1 is the same across all alternatives, even the opt out, then you actually mean that x1 is a scenario variable, not an attribute.

For example, weather (sunny/rainy) is a scenario variable that is not an attribute of a specific alternative. Scenario variables in an unlabelled experiment need to be added to all alternatives in the form of an interaction as mentioned in my previous reply.

Michiel