Skip to: Main Content / Navigation

  • Facebook
  • Twitter
  • LinkedIn
  • Add This

All Forums > Concept Research

Brand and Price Experiment with discrete choice Analysis


for my master thesis I conduct a choice experiment. A consumer has to choose one or none out of 5 branded matching dishes and I want to analyse the data with latent gold choice 4.0. For each brand I'm planning 3 price level. That means 3*3*3*3*3=243 possible combinations of alternatives, is that right? How can I find out which sub sets are the best for the experiment and how many? Another question is, how many respondents do I need for acceptable results,I thought of 300. Problem is, that I can ask each resondent only one choice task.

Thanks for help..

Greetings, Sarah

Brand and Price Experiment with discrete choice Analysis

Because you stated you could only use only one choice task, I think you are wondering about the conjoint method, in which case you would present a ranking task rather than a choose-one task. Correct me if I am wrong.

If I am right about what you propose, you would want to do that conjoint with what is called a "balanced orthogonal" array of choices. Meaning that the combinations of levels of your product features are set up such that no pair is correlated with each other, and the set to be presented is much smaller than the count of all possible combinations.

Come to think of it, even if you do discrete choice and many tasks you will still need an orthogonal array of choices.

It seems to me that what you are describing is 15 discrete levels of price and 5 levels of one product attribute, which we might call "brand".

I don't think it will work as described even in principle, because if the 15 levels really are discrete then they will be correlated with the plates, which will violate orthogonality. Which will bias the estimates of the impact of the attributes.

But presenting only one attribute, price, at 3 different levels for each of 5 dishes (which implies a second attribute, brand or design) is definitely a non-standard application. 15 levels of one attribute and 5 levels of another will make a really big array.

Are you sure that you want to test so many prices for so few other attributes?

Do you really need 15 discrete price levels? Can't you have the 5 dishes share 3 possible price levels? Or even more but not 15?
What is the ultimate goal?

Brand and Price Experiment with discrete choice Analysis

It is very Good reply to this