Millisecond Forums » Millisecond Forums » Inquisit 5 » A question on randomization across groups

A question on randomization across groups

michaelwaltrip — Fri, 13 Dec 2019 08:35:26 GMT

Hi. I have an experiment where, among other things, participants go through the flow in the screenshot below. They're randomized into one of two group (group one gets the "control manipulation" and group 3 gets the "treatment manipulation". Both groups then do the dualtaskadaptivenback.iqx).

When creating the WebScript I answered the question "Choose how the group ID should be determined" by selecting "random selection with replacement". Then for "Enter the total number of groups (if any) in the experiment" I entered 2.

My problem/confusion: In the data files for manipulationcontrolgroup.iqx I see participants with "1"s and "2"s in the group column, not just 1s as the screenshot indicates would happen.. This is also the case for the manipulationtreatmentgroup.iqx. What is happening here? I was under the impression that all participants in group 1 would go into the manipulationcontrolgroup.iqx and all participants in group 2 would be in manipulationtreatmentgroup.iqx. For clarification, participants come to Inquisit from an external participant pool (sonasystems) so their subjectID is predetermined.

If I look in the datafiles I am able to cross reference who went into what group based on the subject's but that is significantly more work than a simple 1 or 2 for group identification. Is there something simple I'm missing?

Thanks

RE: A question on randomization across groups

michaelwaltrip — Fri, 13 Dec 2019 08:35:26 GMT

Ahh, that clears it up. Thanks!

The multiple opportunities to randomize participants is what confused me, I guess. When I responded to the questions on the website indicating that there were 2 groups and they were to be assigned randomly with replacement, I thought I was giving directions for how Inquisit should handle the randomization in the actual script.

RE: A question on randomization across groups

Dave — Fri, 13 Dec 2019 08:21:02 GMT

[quote]

[b]michaelwaltrip - 12/13/2019[/b]

... it looks like my screenshot isn't showing up anymore. The content that the screenshot was showing is the script below. Sorry for the all the issues with my post, hah.

<batch>
/ subjects = (1 of 2)
/ groupassignment = random
/ file = "manipulationcontrolgroup.iqx"
/ file = "dualtaskadaptivenback.iqx"
</batch>

<batch>
/ subjects = (2 of 2)
/ groupassignment = random
/ file = "manipulationtreatmentgroup.iqx"
/ file = "dualtaskadaptivenback.iqx"

[/quote]

The problem here is that you're not assigning your groups based on the group ID / group number.

If you wish to assign based on the group number according to the settings you've taken in your web experiment, you need to set /groupassignment = groupnumber, not /groupassignment = random.

https://www.millisecond.com/support/docs/v5/html/language/attributes/groupassignment.htm

RE: A question on randomization across groups

michaelwaltrip — Fri, 13 Dec 2019 08:07:23 GMT

RE: A question on randomization across groups

michaelwaltrip — Fri, 13 Dec 2019 07:49:46 GMT

I'm not seeing a way to edit my previous post, but I made a typo.

The correct first paragraph should read:

Hi. I have an experiment where, among other things, participants go through the flow in the screenshot below. They're randomized into one of two group (group one gets the "control manipulation" and group 2 gets the "treatment manipulation". Both groups then do the dualtaskadaptivenback.iqx).