When designing my CBC experiment, I carefully select the levels for each attribute respecting these considerations that are often mentioned across studies:
- the number of levels should be similar across attributes (minimum of 3, ideally 5).
- the range of your levels per attribute should include the different values that you want to observe in your study for that attribute
- the range of your levels should not go beyond the realistic values of the attribute
However I don't find any recommendation on the selection of the values within that range. Let us say I have the price attribute and my product can range between 10€ and 50€. If I want to limit the number of levels per attribute to three, can I use 10€, 20€ and 50€? Or should I use a more linear evolution of the price like 10€, 30€ and 50€? Would the first evolution create a bias? Do you know documentation/paper that deal on that topic?