choice-metrics.com

Hi,
I have an experiment design that includes 2 blocks of 6 choice cards each.
Unfortunately the survey programm I am using, while allowing to create two blocks of questions which are randomly assigned to respondants, does not allow to introduce random order of questions within these blocks. I am trying to avoid as much as possible to present choice cards in the same order from respondant to respondant, to avoid learning or fatigue effect biases.
So I was thinking, why couldn't I just skip the blocking part, and create a "pool" of 16 choice cards, and then each respondants would be assigned a random draw of 6 choice card out of those 12 ? This does not seem to be a common practice, but it does not sound bad. I understand the blocking algorithm in Ngene tries to have a good level of attribute representation within the blocks, so as to present each of them equally.
I have not yet found literature discussing this random pooling method. Do you think it would be bad practice?
Thanks

Randomly selecting a subset of choice tasks from an experimental design is actually rather common. It is the default way of giving choice tasks to respondents in the SurveyEngine survey tool I believe. Most survey instruments that are suitable for choice experiments can randomly select choice tasks I think, not all can actually handle blocking by default. There is nothing wrong with that except that you have no control over what combinations of choice tasks are given to respondents. So it could happen that one respondent is given choice tasks with all low price levels. The main reason for blocking is that you have more control over attribute level balance for each respondent. But as you mention, perhaps learning and fatigue effects are stronger than the impact of attribute level balance (which is likely small).

Of course in some situations you may want to have fixed blocks, for example when you are making comparisons across different experimental setups or when you have scenario variables, but in most other cases randomly selecting choice tasks from a pool is perfectly acceptable.

Michiel

Hello,
So I will proceed as such. Thanks for your quick answer.