Page 1 of 1

Main- and interaction effects with more than two levels

PostPosted: Tue Jun 16, 2020 10:51 pm
by kamarci
Dear all,

first of all I want to apologize because I am not very familiar with the topic of choice experiments and constructing designs for them. Nonetheless, I have to figure out which software is best suited task of constructing a choice experiment design. During my literature research I stumbled across a statement concerning Ngene by Burgess et al. (2015, p. 335): "Ngene cannot be used to construct a design for estimating main effects plus interactions when any attribute has more than two levels."

Is that statement true? Because the scenario of wanting to estimate main as well as interaction effects of attributes with more than two levels is one I would believe to be very common. I already looked upon the manual for Ngene version 1.2 but could not find any such restriction there. Was this maybe a bug from earlier versions that has already been fixed?

I thank you very much in advance,
Marcel

Re: Main- and interaction effects with more than two levels

PostPosted: Wed Jun 17, 2020 10:10 am
by Michiel Bliemer
Hi Marcel,

Ngene can do main affects and interaction affects with any number of levels, and has been able to do this for a long time.

The first version of Ngene, released more than 10 years ago, could account for interaction effects between continuous variables (with any number of levels), but only interaction effects with dummy/effects coded variables if they had only 2 levels. We released an update a long time ago to allow interaction effects of dummy/effects coded variables with any number of levels. But note that there are many interaction effects between dummy/effects coded variables if they have a large number of levels.

For example:

Continuous levels:
U(alt1) = b1 * A[1,2,3] + b2 * B[1,2,3} + b3 * A * B

Dummy coded levels:
U(alt1) = b1.dummy[0|0] * A[1,2,3] + b2.dummy[0|0] * B[1,2,3] + b3 * A.dummy[1] * B.dummy[1] + b4 * A.dummy[2] * B.dummy[1] + b5 * A.dummy[1] * B.dummy[2] + b6 * A.dummy[2] * B.dummy[2]

Continuous and dummy coded levels:
U(alt1) = b1 * A[1,2,3] + b2.dummy[0|0] * B[1,2,3] + b3 * A * B.dummy[1] + b4 * A * B.dummy[2]

Michiel

Re: Main- and interaction effects with more than two levels

PostPosted: Thu Jun 18, 2020 4:48 pm
by kamarci
Thank you very much for the reply! :)