contingency tables larger than 2x2
Hello,
I am new to Bayesian analysis, but I am hoping to use JASP for some analyses of contingency tables. Jamil et al., (2017) has been really helpful with this, however all the examples I can find are for 2x2 contingency tables. Is it still possible to use the Bayesian contingency tables option in JASP with contingency tables that are 2x3 or larger?
Thank you for your help, and for this great piece of software!
M
Comments
Dear mspr,
Yes this is possible. If I recall correctly, the Jamil paper also discussed a data set on professions held by fathers and their sons, and that was a pretty large table.
Cheers,
E.J.
Thank you!
Dear mspr:
I am also a new to JASP and i have the same question with you. Can you tell me the paper's name about Jamil et al., (2017)？ thanks！
aha，maybe i find it, lol
the paper is not involved the bayesian contingency table ? I only see the example about correlation, t-test, and ANOVA
https://link.springer.com/article/10.3758/s13428-016-0739-8
Jamil, T., Ly, A., Morey, R. D., Love, J., Marsman, M., & Wagenmakers, E.-J. (2017). Default "Gunel and Dickey" Bayes factors for contingency tables. Behavior Research Methods, 49, 638-652.
Thanks!
Dear EJ:
Thank you for you nice answer yesterday. But now i have a new problem after reading this paper. As referred in this paper, the indicator of the evidence category occurs twice. For example, In table2, 10<BF12<30 means strong evidence , and in figure.1, 2.x<log BF10<3.x means strong evidence. However, in the example 3 the author provided, they obtained a log BF10 =23.03, but they explain it as a strong evidence, that' why? or just a mistake in writing
I checked. The numerical value is correct, so the log(BF) = 23, but in the Jeffreys categorization scheme this is of course "extreme" evidence (we have to take the exponent)
E.J.