choice-metrics.com

Hi everyone its a simple and silly question i was wandering about.

While generating the designs pivot around a reference alternative as (reference alternative, alternative A, alternative B) to check the dominance should the probability of reference also be looked at or the probability of A and B can only be checked? Any reccomendations

You will also need to check dominance with the reference alternative, you do not want the reference alternative to be dominant (e.g. highest quality, lowest price).

Michiel

Thank you for the reply, i am currently designing pivot designs where i have three population segments and am using the fisher(heterogeneous pivot design) property to get three designs, in doing so its difficult to get Utility balance design using b-error. Rather when i design three designs separately without using the fisher property (individually for each segment) and weight i can generate the designs with Utility balance, can three separate designs be constructed without considering weights or fisher property? And what if reference alternative is dominated by other?

1. Yes you can create separate designs, no problem
2. You do not want utility balance, we know that utility balance is bad (e.g. if there is a 50-50% choice probability people are indifferent and randomly choose), optimal choice probabilities are always such that one alternative is somewhat more attractive than another alternative (without being dominant)
3. It is fine if an alternative is dominated if you have 3 alternatives, as long as none of the alternatives are dominant

Michiel

Thank you very much for the reply... 50-50 is undesirable but 30, 70 seems fine though and yes your reply helps very much thank you ....