Sample size justification Bayesian Analysis with BF (2x2 contingency table)
Hello,
I am still wondering about how to justify the sample size for a publication when we use Bayesian analyses. I know that the concept of power is a frequentist concept, and therefore, a power analysis does not seem appropriate. I know that there are attempts at adaptation, notably BFDA, but it is not yet available for all tests. For example, in my case of a Bayesian analysis of a 2x2 contingency table, how can I plan and justify a sample size? However, we can read 'strictly speaking there is no Bayesian need to pre-specify sample size at all (e.g., Berger & Wolpert, 1988).' (van Doorn et al., 2021). So, can I simply dispense with such a justification?
Thank you for your help.
Comments
You could dispense with it, but you might need to appease the reviewers. Maybe this is useful? https://arxiv.org/abs/2406.19940
EJ
Distinct from the issue of statistical power is the question of whether you can demonstrate that you didn't capitalize on chance by halting data collection as soon as you found a sufficiently-extreme Bayes factor to warrant a clear inferential statistical decision. See https://link.springer.com/article/10.3758/s13428-021-01618-1
R