post-pilot parameters

[Joy_Lawrence]

Thank you so much for these suggestions.They were very helpful.
Is there any necessity to have Blocks? For example, if we have 48 choice tasks, and we want every respondent to get 8 choice tasks each, so we mention from 1 upto 8 choice tasks in SE. Will it select 8 choice tasks from a pool of 48, randomly for each person even if we do not mention 'blocks' at all in the design code?
eg. With a sample size of say 900 and with 6 blocks, nearly 150 respondents get the same set of 8 choice tasks, so whether that could be a possible reason behind the error in the design not having enough case-wise alternative specific variation leading to non convergence.

[Michiel Bliemer]

Blocks are not needed. Selecting 8 choice tasks randomly from the 48 rows is also fine, or even cycling through the design by giving the first 8 to respondent 1, the next 8 to respondent 2, etc, and then starting again at the beginning of the design. I believe that SurveyEngine by default cycles through the design but perhaps it random selects, I am not sure. But you can see in your data how the rows were picked from the design when you sort the data by respondent ID.

It is perfectly fine to give 150 respondents the same 8 choice tasks. This is in fact very common, and not a reason for not being able to estimate the model.

Michiel