Compare two proportions with a Bayesian analysis

kev_bague

Hi,

I'm conducting an experiment in which participants receive either an object A or an object B. Next, we ask participants if they want to trade their objects with the alternative object (i.e., object B for participants receiving object A, and Object A for participants receiving object B). I want to compare the proportion or frequency of participants that refuse to trade A with B, with the proportion or frequency of participants that accept to trade B with A.

Is it possible to conduct such an analysis in JASP? Especially, a Bayesian analysis?

Such an analysis looks like an A/B test, except that I want to compare non-occurrence in group A with occurrence in group B. 

Thanks

andersony3k

You do it the same way as described here https://forum.cogsci.nl/discussion/8559/comparing-two-proportions-within-contingency-table-for-small-sample-size#latest except that you select Frequencies, Bayesian, Contingency Tables.

kev_bague

Thanks @andersony3k, I see. However, I have still some difficulties in which such an analysis allow to target failure-groupA vs success-groupB. For me Chi-squared test or Fisher's exact test compute is the proportion of acceptance is different between A and B. But the P-value or Bayes Factor do not result from the assessment of the direct cross-tabulation comparison (no-A ; yes-B)

andersony3k

RE:  "Chi-squared test or Fisher's exact test compute is the proportion of acceptance is different between A and B." It compares the proportion of acceptance between the Initial_A and the Initial_B if you setup your data file that way . . .

INITIAL_OBJECT; ACCEPT_AN_EXCHANGE

A; No

A; No

B; Yes

B; No

But if you want to test the proportions you're interested in, you setup and code *the same data differently* (as indicated in the other thread), thereby comparing the *proportion of Initial_A who stay with A* to the *proportion of Initial_B who switch to A* . . .

INITIAL_OBJECT; ULTIMATELY_GO_WITH_OBJECT_A

A; Yes

A; Yes

B; Yes

B; No

Thus, the P-value or Bayes Factor do indeed "result from the assessment of the direct cross-tabulation comparison," and the particular cross-tabulation comparison depends on how your code your data.