Page randomisation disturbs normal flow of a sequence of choice experiments

[Carla]

Hi,

I have a survey with 3 different choice experiments:

• CE_A and CE_B that are allocated to group 1 of participants (after a filter question). Participants first go through a page of explanation about CE_A + CE_A and then through a page of explanation about CE_B and then CE_B (CE_A and CE_B both comprises 15 choice sets so I have made 3 blocks of 5 choice set in each CE.

• CE_C that is allocated to group 2 of participants (after the same filter question as previous). CE_C comprises 9 choice sets. I made no blocks.

So far, everything goes fine. But when I select the CE pages (for all 3 CEs) for randomisation and for them to be presented in different order to participants, I face a few problems:

• I get 5 choice sets for CE_A but only one for CE_B when I should be getting 5 for each of them
• I get a random number of choice sets for CE_C instead of 9
• Choice sets are mixed up and choice sets from CE_C can appear in CE_A or CE_B

Do you know how I could fix the problem?

Best wishes,

Carla

[Lutz]

This is difficult to answer without seeing your survey. Let’s first establish what you are trying to achieve. Is the following correct?

1. The respondent answers a filter question. Based on the answer, they will be sorted into group 1 or 2. Group 1 answers choice experiments A and B (and not C), group2 answers choice experiment C (and neither A nor B).
2. A respondent in group 1 answers one block of 5 choice tasks from experiment A and one block of 5 choice tasks from experiment B - 10 choice tasks in total.
3. A respondent in group 2 answers all 9 choice tasks from experiment C.

Now a few questions:

1. Apparently, you would like to randomise the order of choice questions. Am I right to assume that you have one choice task per page? It seems that you have put all pages with choice tasks into one randomisation group. Based on what I know, it would seem more appropriate to create one randomisation group per choice experiments in order to shuffle the pages for experiment A among each other, then the pages for experiment B among each other and the pages for experiment C among each other. Maybe this would already fix your problem.
2. Otherwise, how do you achieve the separation into group 1 and 2? Do you use a branch? Do you also uses branches to join everyone back together for the remainder of the survey? Are the branch elements on the same page as choice tasks? What happens if you put the branches onto separate pages?
3. How do assign the blocks and how do you select the design rows for the individual blocks?

[Lutz]

Let’s look at a simple example. My example survey has a filter question at the beginning with 3 options. After that 3 DCEs with 3 choice tasks each (a lot of threes ). Here is the list of pages:

If you pick the first option in the filter question you answer DCE A and then DCE B. You do not see DCE C.

If you choose the second option then you jump directly to DCE B and also skip DCE C (I think this option does not exist in your survey).

Finally, if you pick the third option you jump directly to DCE C.

Let’s look at the filter question:

Based on this question, I used branches to jump to the correct pages:

The intro pages show some text and the choice task pages show one choice task each. In my example, I am showing design rows 1, 2, and 3. You would use your block assignment here.

Importantly, on a separate page after the end of DCE B, I put another branch element to skip DCE C.

I have also put such a branch page after the end of DCE C although this is not strictly necessary in this example.

For the randomisation, open the Options dialog and create a separate randomisation group for each set of choice task pages. This means that the pages will be randomised among each group, but groups will not be mixed.