Designing Fractional Factorial Design Choice Experiment

[#1]

Hi, 

I am using JMP to create my choice profile for the discrete choice experiment. I have two attributes with 3 levels and four attributes with 4 levels. This makes the full factorial design to have 144 possible combinations. However, this one cannot include all options in the survey as it can get tedious for the respondents to answer. For this reason, I want to design my experiment using the fractional factorial method where JMP will choose a subset from the overall combination. I wanted to ask how do we create a choice design based on fractional factorial design?  Are there any important things to keep in mind while JMP chooses the susbet? How do we know JMP makes the best choice of subset?

[#2]

Hi @rm7399,

Welcome to the Community !

I believe we may need additional informations to help you more precisely with your problem.

Are you are using the Custom Design platform (or Classical Design, Full Factorial) to create your experimental plan ?

Or if you are really doing a choice experiment, which means several responders have to choose the best product profile from a set of possible profiles (products) that are proposed in various choice sets, are you using Choice designs (DoE -> Consumer Studies -> Choice Design) ? 

The goal of these two types of design (custom design/fractional factorial or choice design) will be quite different :

• In the Custom Design, you can adjust the number of experiments you run by specifying which terms you want to estimate in your model. Maybe as a first step you could only screen main effects, which should give you a recommended number of run of 32 with the factors you have specified (only categorical, see screenshot "Custom-design_categorical").
• In the Choice design, there are more elements to specify (than just the factors, see screenshot "Design-choice_parameters"), that will directly impact the number of experiments to run : you will have to account for example, the number of respondents, of choice sets, profiles per choice sets, etc... Objective will not be the same, it will be to choose the most optimal product profile among different profiles mixed and evaluated in different choice sets. Evaluation is done by several respondents, and with several different groups of profiles/products, whereas the fractional factorial design mentioned above should have the (32) different products evaluated in the same time (unless you specify a blocking variable like day to split the evaluations on several days).  

If you can explain a little more about your goal (without sharing any confidential/private/sensitive data), that would be much easier to figure out which design type best suits your study.

Sorry for the delay, I hope this first answer will help you,

[#3]

Hi Victor,

Thank you for getting back to me. This answer definitely helps. I am mentioning below more about my study. It would be great if you could help me with this. 

I am conducting choice experiments with my respondents. As you mentioned, I have used Choice designs (DoE -> Consumer Studies -> Choice Design) to create my choice profiles. 

In the choice design, I did mention more information like the number of respondents per survey (200), number of choice sets (8), profiles per choice sets(2), number of attributes that can change within a choice set(6) number of surveys (3). 

To give you a background, I am conducting choice experiments with respondents where I will present them a set of choice profiles out of which they would need to select their preferred one. I have 6 factors (traits), 2 factors with 3 levels and 4 factors with 2 level. The total possible combination with this set would be 144. However, we cannot include 144 combinations in the survey.  For that reason, I decided to mention 3 surveys so that I can randomly assign one block/survey to one participant. 

JMP gave me 3 surveys with the possible combinations. Can I say this is a fractional factorial design because I got 48 possible combinations (a subset of 144 combinations)? Is this the correct way to do it? Also, how can I check the d-efficiency level? I have read research papers on choice experiments which have used JMP and they often mention the d-efficiency score. I am not able to figure out how can I check that. 

Thanks 

[#4 · Accepted solution]

Hi @rm7399,

Choice designs may be seen as a specific subset of Fractional Factorial Design, with only categorical factors, a discrete choice response and often some constraints (for example the number of factors/attributes that can change in a subset of experiments/choice set). As you mention, you are doing only a subset of all possible combinations, since you're interested only in main effects.

You can perhaps watch this video, showing a step-by-step presentation of Choice Designs creation and analysis, it may help you in the process of creating and analysis you design: Case Studies on Designing and Analysing Discrete Choice Experiments Using JMP® - JMP User Community

One way to look at D-efficiency score would be to go to DoE -> Design Diagnostics (jmp.com) -> Evaluate Design, and enter your categorical factors only. You will have at the end of the report the different efficiencies scores. Efficiencies scores are interesting information, but more valuable when taken and used in the comparison of several designs.

I hope these informations will help you,