Groups within groups
Daniel Nettle
I have a query about Lioness and I apologize if it is extremely elementary. I have not started using Lioness yet, I want to establish if it will be easy to do the thing I need.
My experiment has the following structure:
Participants arrive into a lobby, and once 14 are present, they are grouped up into a session (I want to allow for the possibility of many such sessions running at a time, i.e. another 14 show up at the lobby, they are then grouped and go off to start).
Within a session, each round, pairs are formed at random, based on a colour which has been assigned to each player. So the actual interaction group size is 2, even though the session group size is 14. The pairs are reformed from the same group of 14 for each of many rounds.
I hope that is clear - if someone can just give me a pointer on how easy it will be (i) to have this pairs within groups situation; and (ii) to match up pairs based on a colour variable that is assigned prior to round 1.
Thanks very much!
in...@lioness-lab.org
Hi Daniel,
Thanks for your question. The short answer is yes. The long answer is this can be tricky, especially when the participants are recruited from online platforms like Prolific.
Your design seems to require groups of 14 participants to move in synchrony through a number of rounds. Once 14 participants have gathered in the lobby (usually best placed after instructions and perhaps control questions), a group will be formed.
You can manually set up the pairs for each round by using the participants’ variable “subjectNr”. Once the group is formed, each participant will get a subjectNr, in your case ranging from 1 to 14. You could assign “yellow” to participants with subjectNr <= 7 and “blue” to participants with subjectNr >=8. To get yellow-blue pairs, you can assign partners in round t as a participant’s own subjectNr plus or minus (7 + t) %% 7.
You can pre-determine the partners for each round and save them in a matrix. In each round, you can wait until each group member has made their decision, and calculate the results based on that matrix (for each player, get the subjectNr of the current partner and retrieve that partner’s decision in the current round).
The tricky part here is that (regardless of the platform in which you program the experiment), the group will proceed at the pace of the slowest member. In a physical lab the experimenter can usually just go to the booth/cublicle of the slowest player and ask them to continue. Online, a slow pace might well lead to distraction cascading through the group, and participants dropping out altogether. The default in LIONESS is that upon participant dropout, the experiment proceeds with a group of reduced size.
With a “static” predetermined matrix of interaction partners, you might have to program in “fallback” decisions should a partner have dropped out (that is, in the current round, no decision has been recorded for the group mate with the subjectNr of the current partner). In LIONESS replacing dropouts with fresh participants is not easy (it will not yield clean data anyway).
Perhaps there are more clever ways to deal with this, but this is what came to my mind first. Should this look promising, I’d be happy to try and set up a quick prototype. There might be other platforms (e.g., oTree?) that deal with this in a more sophisticated way.
Hope this is somehow helpful!
Best wishes,
Lucas
--
You received this message because you are subscribed to the Google Groups "LIONESS Lab help and discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lioness-lab...@googlegroups.com.
To view this discussion, visit https://groups.google.com/d/msgid/lioness-lab/eb1190e9-4d77-46ad-9da3-944e2ea55d37n%40googlegroups.com.