Is there a reason why you are using the modified Federov algorithm? This algorithm relaxes attribute level balance and may not use all attribute levels if they are not needed (in your case, it is more efficient to use only 10 and 30 minutes). If you want all levels to appear equally in the design you should either use the default swapping algorithm (so simply remove ;alg = mfederov) or you should imposed attribute level constraints, e.g. ... + b3 *time[10,20,30](3-5,3-5,3-5) + ..., which means that each level for time should appear 3 to 5 times across the 12 choice tasks.
Thank you so much for your reply. I read the topic "A way of decreasing D-error", and you suggested that
Aspects that decrease D-error:
* More choice tasks per respondent
* More alternatives
* Wider range for quantitative variables such as pr
* using the modified Federov algorithm to generate the design, which removes the attribute level balance constraint
then I tried to use the modified Federov algorithm to decrease D-error, but I do not understand the violation of level balance.
From your recommendations, I will remove ;alg = mfederov from the syntax.
Yes the modified Federov algorithm removes the attribute level balance constraint and therefore is able to achieve a lower D-error. If you want to have attribute level balance, as you indicate, it is best to use the default swapping algorithm.