Millisecond Forums » Millisecond Forums » Inquisit 4 » Random assignment to groups/conditions + counterbalancing

Random assignment to groups/conditions + counterbalancing

nmwyale — Wed, 24 Aug 2016 14:46:35 GMT

I am conducting a study in which participants need to be assigned to one of 8 conditions. Within each of these conditions, I need to counterbalance the order of presentation of two items (with two levels each). I am trying to determine the best way of accomplishing this. As of now, I have created 32 experiments.

For example:

<expt Condition1>
/ preinstructions = (intropage)
/ groupassignment = random
/ subjects = (1 of 32)
/ blocks = [1=InfoTarget1; 2=Faceblock; 3=Fillerblock; 4=Presentation1CounterbalanceOrder1; 5=Presentation2CounterbalanceOrder1]
</expt>

I have two main questions:

1. If all I was looking to do was, indeed, to randomly assign participants to one of 32 conditions, would this (above) work? And if so, is it correct that for a Web Script, I could select "subject ids should be generated via URL parameter" (participants will be randomly assigned to a subjectid in Qualtrics, and this will be included in the Inquisit Link) and "group id should be determined via Random Selection Without Replacement"?

2. Is there a better way to do the counterbalancing within conditions so that I do not need to create 32 distinct conditions (and could instead include 8 conditions with 2 order variables with 2 levels each counterbalanced within each condition)?

Thank you in advance for your help!

-Natalie

RE: Random assignment to groups/conditions + counterbalancing

Dave — Wed, 24 Aug 2016 14:46:35 GMT

Qualtrics generates a unique participant ID automatically -- most people will want to share that ID with Inquisit Web (by having Qualtrics forward its ID to the Inquisit launch page via a URL query parameter), which obliterates the need for having Qualtrics generate some random numerical value.

See the "How to interoperate Inquisit Web with online survey packages" topic in the documentation for information re. the Inquisit-side of things, and see the links to the pertinent topics in the Qualtrics documentation in this post: https://www.millisecond.com/forums/FindPost12409.aspx

RE: Random assignment to groups/conditions + counterbalancing

nmwyale — Wed, 24 Aug 2016 14:23:36 GMT

Thank you, Dave. This is quite helpful. I will change "random" to "groupnumber," have groupnumber generated randomly, and use the group = ((groupid - 1) % groupcount) + 1 formula to determine which group a given participant was assigned to (1, 2, 3, ... 30, 31, or 32).

To clarify: When I said "participants will be randomly assigned to a subjectid in Qualtrics," what I meant was that I have added a random number generator to Qualtrics, which will assign a number between 1,000,000,000 and 9,999,999,999 labeled "subjectid." Is it okay to use that rather than "Have Qualtrics forward its ID via a query parameter and have the Inquisit Web experiment retrieve its value and use it as subject id as you planned"? (I'm a bit confused about the latter option.)

Thanks again.

RE: Random assignment to groups/conditions + counterbalancing

Dave — Wed, 24 Aug 2016 12:54:31 GMT

#1: Qualtrics-generated IDs are not numerical (but alphanumerical), and hence are unsuitable for condition assignment. You should
(a) Have Qualtrics forward its ID via a query parameter and have the Inquisit Web experiment retrieve its value and use it as subject id as you planned, but
(b) Change your <expt> /groupassignment method to groupnumber instead of random:

<expt Condition1>
/ preinstructions = (intropage)
/ groupassignment = groupnumber
/ subjects = (1 of 32)
/ blocks = [1=InfoTarget1; 2=Faceblock; 3=Fillerblock; 4=Presentation1CounterbalanceOrder1; 5=Presentation2CounterbalanceOrder1]
</expt>

In your web experiment's settings select the groupnumbers to be generated either randomly or sequentially (those are simply choices from a dropdown menu when you upload and configure your script to your Inquisit Web account).

The above will ensure that (1) a participant will be randomly assigned to one of the 32 conditions, and (b) that there's a relationship between the administered condition and the generated groupnumber (i.e., you can determine which condition was administered by looking at the groupnumber): https://www.millisecond.com/forums/Topic13856.aspx

#2: No, I don't think there is a better way in this case. Your counterbalancing factors are between-subjects just like the 8-level main factor. I.e., you end up with 8 x 2 x 2 = 32 conditions.

Hope this helps.