Appropriate analysis question

#1
Hey all,

I have a question I’m hoping you might help me with.

The research question is whether participants would distribute a virtual currency equally, give more of it to themselves or to their partner (okay, the research question is a bit more complicated than that but this is the general structure).

In the experiment, participants are presented with a series of options on how to distribute currency:
(IV) 3 payouts for self: High, medium, low
(IV) 3 payouts for partner: High, medium, low

Each trial, they get two options for distribution and each option has a payout for self and a payout for partner. For example:
Option A: Medium self payout + low partner payout
Option B: Medium self payout + medium partner payout

The dependent variable is participants’ choices of option A or B (i.e. the payouts for themselves and their partner).

Each combination between self and partner payout is presented as an option an equal number of times (8) and each option pair is presented once. So on the example of High self + High partner payout the choice look like:
Trial 1: Option A: High self + High partner Option B: High self + Medium partner
Trial 2: Option A: High self + High partner Option B: High self + Low partner
Trial 3: Option A: High self + High partner Option B: Medium self + High partner
Trial 4: Option A: High self + High partner Option B: Medium self + Medium partner
Trial 5: Option A: High self + High partner Option B: Medium self + Low partner
Trial 6: Option A: High self + High partner Option B: Low Self + High Partner
Trial 7: Option A: High self + High partner Option B: Low Self + Medium Partner
Trial 8: Option A: High self + High partner Option B: Low Self + Low Partner

I was wondering what would be an appropriate analysis to see participants’ preferences for self and other payouts (e.g. do they prefer medium-self and medium-partner or high-self and medium partner)?

Please let me know if there’s any other information about the design I can provide that might help.
Thanks in advance!
 

Karabiner

TS Contributor
#2
So each participant is presented with all 36 possible comparisons between different combinations?

And how large is your sample size?

With kind regards

Karabiner
 
#3
Hi Karabiner,

thank you for the reply.

Yes, each participant goes through all the 36 possible combinations (36 trials in total per participant).

I haven't decided on a sample size yet. As a ballpark I was thinking around 60, but I'll run it online so I have some flexibility in increasing the sample size if there's a good reason for it.

(On a sidenote - would a binary logistic regression make sense in my case?).

Best,
Anteater
 

Karabiner

TS Contributor
#4
I am not sure what exactely you want to find out. "whether participants would distribute
a virtual currency equally, give more of it to themselves or to their partner (...) to see participants’
preferences for self and other payouts e.g. do they prefer medium-self and medium-partner or
high-self and medium partner" is a bit confusing. Why such a complicated design? Could
you tell a bit more about the idea behind it?

With kind regards

Karabiner
 
#5
Hi Karabiner,

thanks for the reply. Reading back, I don't think I described the study completely accurately.

I didn't mention that in between the choices, participants do a simple task to which they have an equal contribution but they can't complete on their own. There is some research showing that when this is the case, people give their own and their partner's contributions equal weight. I'm interested if a similar thing happens with rewards.

The word "equally“ was misleading. I think a better description would be whether, when sharing rewards, participants try to maximize their own gain regardless of the gain of the partner, maximize the partners' gain regardless of their own or try to maximize the total gain (theirs+partner's). The maximization of total gain would mean that sometimes they choose options that are not optimal for themselves but are for their partner.

I hope this makes the reasoning a little bit clearer. If not – apologies. Please let me know if I should expand on this.

Best,

Anteater
 

Karabiner

TS Contributor
#6
I didn't mention that in between the choices, participants do a simple task to which they have an equal contribution but they can't complete on their own. There is some research showing that when this is the case, people give their own and their partner's contributions equal weight. I'm interested if a similar thing happens with rewards.
So they do 36 comparisons, and between comparisons they do that task? Or, they do some comparisons, do the task, and then do the other comparisons? Or they do 36 comparisons, then the task, then again the 36 comparisons?
when sharing rewards, participants try to maximize their own gain regardless of the gain of the partner, maximize the partners' gain regardless of their own or try to maximize the total gain (theirs+partner's). The maximization of total gain would mean that sometimes they choose options that are not optimal for themselves but are for their partner.
So you could code each comparison with regard to which of these goals was achieved?
I still don't really see why there's such a complicated design. Is it modeled after some
other studies in the field?

With kind regards

Karabiner
 
#7
Hi Karabiner,

thanks for the reply.

To answer your question, they do the task and then they’re asked to make a choice about the payoffs and then they do the task again and and then make a payoff choice and so on.

I did a bit of thinking and simplified the design. In short, instead of presenting the participants with two choices every trial, they’ll be presented with one possible combination of payoffs (e.g. High self-medium partner) and asked to rate on a slider how much they want it as an outcome at the end of the experiment.

Then it’s pretty straightforward to just present them with all the payout combinations one by one and get their mean preference for each of those.

If I’m missing something, please let me know.

Best,

Anteater