Randomizing blocks with non-identical probabilities | XM Community
Solved

Randomizing blocks with non-identical probabilities

  • 12 July 2018
  • 3 replies
  • 277 views
L
I have an experiment with several treatments and one control group. There are three scenarios in the experiment, and every scenario has several treatments. I have constructed the survey so that every scenario corresponds to one block, and the treatments are then administered as randomly assigned question (with only descriptive text) within the blocks.

This works well, except for the fact that I want to make sure that individuals who are assigned to the control group only get the control in all three scenarios. My initial solution to this problem was to create a control block, in addition to the three scenario blocks where treatments are administered. However, since Qualtrics randomizes with even probabilities across blocks, this would lead to 1/4 of the participants being assigned the control scenario. Ideally, I would like to assign 1/10 of the respondents to the control group, since my most interesting comparisons for this experiment is across different treatment groups. The control is there simply to serve as a benchmark.

Is there any way to randomly assign respondents to different blocks with different probabilities?

Many thanks,
icon

Best answer by mattyb513 12 July 2018, 12:38

View original

3 replies

mattyb513
Rank
Userlevel 6
Badge +6
I think you could solve this by adding the scenario blocks multiple times to the randomizer. Basically you would have something like this:

Randomizer:
- control
- Treatment1
- Treatment2
- Treatment3
- Treatment4
- Treatment1
- Treatment2
- Treatment3
- Treatment4

This would, I guess, only be 1/9.

Another solution would be to use the Random number generator. At the top of your survey flow you could assign an Embedded Data field to a random number between 1 & 10 using the ${rand://int/1:10} piped text.

Then have a branch that if that value = 1 to put them to show them the control.

If that number does not equal 1 then put them in a randomizer to see one of the remaining 3 blocks.

Read more about the random number generator here: https://www.qualtrics.com/support/survey-platform/survey-module/editing-questions/piped-text/piped-text-overview/
L
Thanks a lot! The latter approach sounds very promising. I'll give it a go.
rochak_khandelwal
Rank
Userlevel 3
Badge +12

I think you could solve this by adding the scenario blocks multiple times to the randomizer. Basically you would have something like this: Randomizer: - control - Treatment1 - Treatment2 - Treatment3 - Treatment4 - Treatment1 - Treatment2 - Treatment3 - Treatment4 This would, I guess, only be 1/9. Another solution would be to use the Random number generator. At the top of your survey flow you could assign an Embedded Data field to a random number between 1 & 10 using the ${rand://int/1:10} piped text. Then have a branch that if that value = 1 to put them to show them the control. If that number does not equal 1 then put them in a randomizer to see one of the remaining 3 blocks. Read more about the random number generator here: https://www.qualtrics.com/support/survey-platform/survey-module/editing-questions/piped-text/piped-text-overview/

Hey @mattyb513 - I have a query regarding the second approach you told (using random number generator). How do I make sure that the randomly generated numbers would have uniform distribution?

 

Please suggest.

Thanks.

Leave a Reply


Terms & Conditions