Choice Design

[#1]

Dear All,

I have a question regarding the choice design feature on JMP. I have 4 attributes and I want to make a choice design that each choice consists of 3 attributes instead of 4. I only know how to make that choice design that consists of 4 attributes for each. Anyone would help me, so much appreciate. Thank you

[#2]

Hi @arismnndar,

It sounds like you might be looking for the option "Number of attributes that can change within a choice set," which you will find in the Design Generation panel when you're designing your choice experiment. 

In your case, if you set that number to 3, JMP will design the sets such that 3 attributes will vary in the choice sets with the final attribute held constant.

Here's a link to the documentation on that topic if you need more details. 

I hope this helps!

@julian 

[#3]

Hi Julian, 

Thanks for your response. I tried that way and the result remains with a choice set that consists of 4 attributes rather than 3 attributes.

[#4]

Hi @arismnndar,

I understand better now, you don't just want that attribute to be held constant, you wish that it is not displayed at all. How are you distributing/administering this experiment? This is less about the design generation and more about the generation of the testing apparatus. I think it's going to come down to figuring out a method to help you not pass along the attributes that are held constant, but the best way to do that is going to depend on how you're administering this. 

@julian 

[#5]

Yes, that is true Julian. For example, I have attributes A, B, C, and D. Each attribute has two levels (L1 and L2). I wish to make a design for example; choice 1. A1, B1, C2; A1, B2, C2 and etc. Can I do it through JMP Pro as well?

[#6]

If you do not wish to include D at any point, you can simply not include it in the design generation. But, I don't think that's your intent, correct? Assuming that you do wish to include all your attributes, and for each set have a different attribute held constant (and invisible), what is going to happen after designing your experiment using DOE > Consumer Studies > Choice, is you will get a table of your choice sets, which one of the pair in each row. If you indicate you want only three attributes to vary in a set, at least one of the attributes will be fixed. There isn't a way to suppress the display of the fixed attribute for each choice set (and, it will be a different attribute in each choice set that is held constant). But, I expect you won't have people using JMP to fill out their responses (assuming this is for human participants). So, my question was what software (survey or otherwise) will you be using? I ask because you will need to convey the survey design over to that format, and it's in that translation that you will have control over not including the details of the attribute that is constant. In the most low-tech version of this, you simply do not include the details about the attribute that is held constant when you type in the details of the choice set. Again, it's not a feature of the design, but of the administration. 

That all said, your choice to not include that information is not neutral psychometrically. The absence of that attribute will certainly divert additional attention on the remaining attributes (which might be your intention), but if the omitted attribute has interactions with the other attributes you are robbing yourself of data in which that effect can emerge. You are additionally removing data for your model to estimate the average inertia of the levels of that attribute (at least when you have a no-choice option, perhaps less so with a full forced choice design). I can't speak completely to the statistical implications of that (I haven't done the necessary research to state anything definitive), but if you pursue this I would suggest finding someone expert in choice designs (which unfortunately isn't me) to think through in detail the consequences of this with you. 

[#7]

Hi Julian, 

That's true. Thank you very much for a very good explanation. I got your point. So I would include all attributes for each choice set then. 

[#8]

I think that's probably the best thing to do. Good luck!

[#9]

Hi Julian,

I have a question regarding DOE choice design on JMP. I have designed a questionnaire on JMP DOE choice design with 4 attributes and 3-4 levels for each. I created 10 choices set. However, in the questionnaire website, I add more questions such as age, gender, income, country. I also want to analyse it in JMP. Could you tell me please how do I input those data in JMP and how to analyse it? Thank you