Any suggestion how best to accomplish this? Thanks!

Visit Sawtooth Software Feedback to share your ideas on how we can improve our products.

Any suggestion how best to accomplish this? Thanks!

+1 vote

Thanks for the clarification.

I would generate the 'A' concepts using the standard tools.

Then in Excel double up the concepts for each task and randomly tweak the 'B' concepts (every other task when ordered - identical to A except for the necessary variation).

In the software add in the attribute, for the 'random variation' and then import the new design into the software (so you now have each concept with two variants and an attribute that identifies the variant).

Does this make sense? It depends on my understanding of your original query.

I would generate the 'A' concepts using the standard tools.

Then in Excel double up the concepts for each task and randomly tweak the 'B' concepts (every other task when ordered - identical to A except for the necessary variation).

In the software add in the attribute, for the 'random variation' and then import the new design into the software (so you now have each concept with two variants and an attribute that identifies the variant).

Does this make sense? It depends on my understanding of your original query.

Yes, this makes great sense. What is your opinion on these two options:

1. Exactly as you said, generate a design that has, say, 3 concepts each task, 6 tasks each respondent. Then double the design and add the randomized variants attribute, so now each version has 3 concepts each task, 12 tasks each respondent.

2. Generate a design that has, say, 3 concepts each task, 6 tasks each respondent. Then double the concepts for each task. And add a variants attribute so that the original concepts would have a level 1 and the doubled concepts have a level 2. Therefore there are 6 concepts each task, and 6 tasks each respondent.

1. Exactly as you said, generate a design that has, say, 3 concepts each task, 6 tasks each respondent. Then double the design and add the randomized variants attribute, so now each version has 3 concepts each task, 12 tasks each respondent.

2. Generate a design that has, say, 3 concepts each task, 6 tasks each respondent. Then double the concepts for each task. And add a variants attribute so that the original concepts would have a level 1 and the doubled concepts have a level 2. Therefore there are 6 concepts each task, and 6 tasks each respondent.

As you are effectively doubling the number of concepts I would see if I could get away with two initial concepts per task - doubling up to four. Then aim for 12 tasks. However, this would need the Advanced test to investigate the standard errors. It depends though on the number of other attributes/levels in the design. If you are lucky enough to have few, or can present more in a simplified way, then the cognitive burden is easier and you can aim for more concepts and fewer tasks. Peoples heuristics do change as the number of tasks increases. Which is more realistic (earlier versus later answers) is debatable - however - fewer tasks is likely to be more consistent!

...