choice-metrics.com

Dear Michiel,

I made an design using OOD. Is that ok to run model with interaction effect after I collect data from OOD design without interaction since OOD design can not have interaction?

Thanks,

Wei

You should generally have no problems in estimating interaction effects even if you did not account for them in designing your experiment. Only in rare cases, where the interaction effects are perfectly confounded with one of the main effects, you would not be able to estimate them. You can easily check in Ngene when you inspect the design, click on Correlations and then Interactions. If you notice an interaction of 1 between a main effect and an interaction effect within the same alternative, then that specific interaction effect cannot be estimated.

Michiel

Thanks so much for your help.
Wei