Partial-profile design with prohibition

Question

I m currently design a partial-profile conjoint study, and find something is difficult to set up. So it would be appreciated if anyone could help me.

My study includes 11 attributes. Each attribute has 2 levels which indicate the attribute is available or not. Thus, it looks like below:

Attribute 1: Level 1 - Available; Level 2 - Unavailable
Attribute 2: Level 1 - Available; Level 2 - Unavailable
...
Attribute 11: Level 1 - Available; Level 2 - Unavailable

For each attribute, I want to only show "available" level which is Level 1, so I define null level prohibitions within prohibitions dialogue. I tick Level 2 (Unavailable) for each attribute, but a bit confused to set the number for minimum / maximum null levels per concept in order to meet my needs. (Currently, I set 7 for minimum, and 8 for maximum)

Also, for each concept, I want to always only show 3 attributes per concept, and 4 concepts (plus none option) per screen.

Then I go to "Advanced Design Module" dialogue, and select "Partial-profile CBC Design", and set 1 for "Rotate Attributes into Concepts Starting with", 11 for "and Ending with Attribute" in order to make full rotating attributes. Besides, 3 for "Number of Attributes Shown in each Concept".

BUT, when I press "Preview" button....
 
A warning dialogue pops up: The following preview error occurred: The task could not be generated according to the constraints. Please check the prohibitions.

I really don't know what I did wrong. Also, the warning dialogue pops up when I press "Generate Design" button.

I use SSI Web 6.6.12 to design this study.

My question is how to what things I did wrong, and how to solve this problem by SSI Web 6.6.12. Thanks so much for your help.

Answer 1

Sounds like your design is too prohibitive for our experimental designer (too many restrictions about exactly how many "null" levels are to appear per concept.  If you make the minimum and maximum range somewhat wider (rather than 7 to 8 null levels permitted, maybe something like 5 to 10), then maybe it would work better.  

Maybe it's a study that would be more appropriate for MaxDiff?  Depends on your aims.

Comments

Hi Bryan

Thanks for your reply.

I followed your suggestion, found there was no warning dialogue. But one more question, the concept does not always show only 3 attributes. Some concept shows 1 attribute, some shows 2, and some shows 3. Actually, what I want is all concepts always show 3 attributes. Do you have any suggestion about this issue?

Re the MaxDiff, as I want to use simulator to do "what-if" analysis, thus, I consider to use partial-profile design. Can MaxDiff support "what-if" analysis via simulator?

Thanks again for your kind answer

---

I've found that if you try to always show the same number of "true" levels from a series of binary attributes, you get a deficient design.  So, you have to vary the number of items shown as "true" in order to estimate parameters.  Maybe somebody out there can give a solution for this, but that's what I've seen.

Oh, one solution that occurs to me is to occasionally hold one or more of the items constant "true" across all concepts in the CBC task.  I think that might break up the indeterminacy and allow estimation.  But, this cannot be done automatically using our design program.  You'd need to do some manual "massaging" of the design files to make this happen.

---

Thanks Bryan. I have another idea, if my client wants to have 50-version designs, I would generate more than 50 versions, say 80 versions or 100 versions. Then I export the design. NExt, I select designs which includes only 3 attributes from 80 or 100 versions, and paste them into new CSV file to fulfill the design number of 50-version designs. And use "Import design" function to import the design. So my design will have only 3 attributes. How about this way? Is there any concerns or risks to do so?

Thanks.