Hello DCE experts?
I am designing a discrete choice experiment using Stata's dcreate command.
I have a general question concerning the trade off between obtaining more data points from a smaller number of choice questions and obtaining fewer data points from a greater number of choice questions.
That is, if a researcher needs to create a partial factorial design due to a huge total number of choice sets, one can imagine creating 12 choice questions only, each of which consist of 2 alternatives, and have all participants, let's say 200 participants, solve the same 12 choice questions.
Or, one can imagine another situation in which he or she creates 24 choice questions but divide them into two blocks with each block containing 12 choice questions, and have a half of the participants (i.e., the first 100 participants) solve the first 12 choice questions in block 1 and the other half (i.e., the remaining 100 participants) solve the remaining 12 choice questions in block 2.
Among these two scenarios, which case provide more, accurate information about parameters? In the first scenario, the partial factorial design (i.e., 12 choice questions) covers a smaller space of the full factorial design, which may hamper accurate estimations, but it allows us to obtain more data points per choice question, which helps accurate, efficient estimations, compared to the second scenario.
On the other hand, in the second scenario, the number of data points from per choice question is smaller, but the partial factorial design (i.e., 24 choice questions) covers a greater space of the full factorial design, given that none of the 24 choice questions is the same.
I understand that the answer may vary depending on the numbers of participants, choice sets, and alternatives per choice set, but I wonder if there is a general tendency in this trade-off.
Thank you very much for your interest in this matter.
I am designing a discrete choice experiment using Stata's dcreate command.
I have a general question concerning the trade off between obtaining more data points from a smaller number of choice questions and obtaining fewer data points from a greater number of choice questions.
That is, if a researcher needs to create a partial factorial design due to a huge total number of choice sets, one can imagine creating 12 choice questions only, each of which consist of 2 alternatives, and have all participants, let's say 200 participants, solve the same 12 choice questions.
Or, one can imagine another situation in which he or she creates 24 choice questions but divide them into two blocks with each block containing 12 choice questions, and have a half of the participants (i.e., the first 100 participants) solve the first 12 choice questions in block 1 and the other half (i.e., the remaining 100 participants) solve the remaining 12 choice questions in block 2.
Among these two scenarios, which case provide more, accurate information about parameters? In the first scenario, the partial factorial design (i.e., 12 choice questions) covers a smaller space of the full factorial design, which may hamper accurate estimations, but it allows us to obtain more data points per choice question, which helps accurate, efficient estimations, compared to the second scenario.
On the other hand, in the second scenario, the number of data points from per choice question is smaller, but the partial factorial design (i.e., 24 choice questions) covers a greater space of the full factorial design, given that none of the 24 choice questions is the same.
I understand that the answer may vary depending on the numbers of participants, choice sets, and alternatives per choice set, but I wonder if there is a general tendency in this trade-off.
Thank you very much for your interest in this matter.
