Millisecond Forums

Getting better balance in conditions with group assignment

https://forums.millisecond.com/Topic23860.aspx

By InquisitivePhoenix - 2/12/2018

Hi,
I recently ran an online experiment with two conditions, participants were assigned into one randomly (groupassignment = random). The result was nearly two thirds of participants in one condition and a third in the other. I've been looking at ways to use group assignment to assure more balanced numbers on my second run. The Inquisit docs says the "subjectnumber" assignment is "based on subject number" but how does this work exactly? Can I use it to essentially alternate between the conditions during assignment, rather than it being random?

Thanks!
By Dave - 2/12/2018

InquisitivePhoenix - Monday, February 12, 2018
Hi,
I recently ran an online experiment with two conditions, participants were assigned into one randomly (groupassignment = random). The result was nearly two thirds of participants in one condition and a third in the other. I've been looking at ways to use group assignment to assure more balanced numbers on my second run. The Inquisit docs says the "subjectnumber" assignment is "based on subject number" but how does this work exactly? Can I use it to essentially alternate between the conditions during assignment, rather than it being random?

Thanks!

You can do the following: Set /groupassignment to either subjectnumber or groupnumber (whatever is better in your particular case) and then choose sequential subjectnumber and/or groupnumber generation in your web experiment's settings.

Alternatively, if you know who / how many will be participating beforehand, you can also pre-assign your participants to conditions (half to condition 1, the other half to condition 2). To do this, you'd configure your web experiment to read the groupnumber (used for condition assignment) from a query parameter you can pass in to the launch page. Then, essentially, you'd give half your participants a link that runs the 1st condition

https://research.millisecond.com/yourusername/yourexperiment.web?groupid=1

and the other half a link that fires up the 2nd condition


https://research.millisecond.com/yourusername/yourexperiment.web?groupid=2