choice-metrics.com

Hello,

I am reviewing a paper where authors have used a choice design with food products (unlabeled choice experiments) which have created 15 choice sets which they divided into 2 blocks of 8 and 7 choice sets. Thus, one a group of consumers get 1 block of 8 choice sets while the other group get a block of 7 choice sets. I know that the standard practice is to have blocks of equal number of choice sets, I think. I am wondering if using blocks of different sizes of choice sets is correct/potential implications or consequences, I guess related to statistical efficiency.

Best regards,
Daniele

While it is uncommon to have blocks of unequal size, it is not really a major issue. It is for example also possible that some respondents do not complete all choice tasks and therefore one would obtain a different number of observations from respondents, which is not a problem. The main thing to consider is fatigue, so some respondents were given 7 choice tasks while others were given 8 choice tasks. But you could argue that fatigue effects will not be very different between 7 and 8 (but would be between 7 and 17, for example). Having different block sizes does not effect efficient since efficiency is based on the entire design, in this case 15 choice tasks.

Michiel