choice-metrics.com

Dear Michiel,

Apologies for another question but I just wanted to clarify that manually swapping choice tasks across blocks would only affect attribute level balance, and not identification or efficiency or any other design properties?

While I understand that it is not possible to achieve a high degree of balance within blocks for all attributes with efficient designs, I'm concerned about cases where some levels don't appear even once in each block, since that could lead to potential framing effects. So I'd like to try some manual rearrangements to ensure that each level is seen at least once in each block.

Also, is there any recommended way of doing this or is it just based on trial and error?

Many thanks,
Ashani

Correct. Swapping choice tasks across blocks does not affect any design properties, so you can safely do this. The efficiency of the design remains exactly the same.

There is no recommended way of doing this, indeed just a bit of trial and error.

Michiel

Thank you very much.

Together with colleagues I have been working on a new blocking algorithm that we aim to submit for publication in the Journal of Choice Modelling later this year. I now have a version working in Excel (using Solver to do the optimisation) and it seems to do a better job in achieving attribute level balance.
Would you like me to try it on your design? If so, please provide a link to the file or send it by email.

Michiel