choice-metrics.com

Hi,

I've got two questions which I couldn't find in the Ngene manual:

1. I want to include one or two 'holdout profiles' (profiles that test the model's predictive power) in my Ngene design. How can I do this?
2. I want to estimate the interaction effect between two dummy-coded two-leveled attributes (0 and 1). However, when I run Ngene it tells me that ''Interactions of dummy or effects coded attributes are not currently supported''.

Could you help me please? Thank you in advance!

Yours sincerely,

Ilias

1. Ngene generates designs for estimation only, and does not account for hold out samples. If you want one or two choice tasks as a hold out, you can simply select them from the existing design. If you are worried about what removing those choice tasks from the estimation data will do to the efficiency of your design, you can save the design in Excel, remove the choice tasks, and evaluate in Ngene (using ;eval = ...). Alternatively, you can manually add choice tasks to your design after you have generated an efficient design and use those as a hold out sample. Since they are not used for estimation, it does not matter too much what their efficiency is.

2. Interactions between dummy or effects coded variables can be done by individually specifying each interaction (since interaction between such variables requires estimating multiple parameters). Please refer to p. 126 of the Ngene manual.