No, especially because the question is complicated by the issue of priors. However, in general the same rules apply - bigger samples produce tighter parameter estimates.

BA usually depends on priors.

However, If you able to manage a bigger sample, analysis would be great, even if you have a good informative prior.

Regards,

Abhay

Hi all,

although the question is 2 months old, there seems to be the, i.m.o., most important answer missing, which is a reference to this article:

Schönbrodt, F. D., Wagenmakers, E. J., Zehetleitner, M., & Perugini, M. (2017). Sequential hypothesis testing with Bayes factors: Efficiently testing mean differences. Psychological Methods, 22(2), 322.

Basic message: (also brought up against standard Null-HypothesisTesting) Even if your "t-test" (or any other classic 'standard method') produces a significant result, while having an inconclusive Bayes Factor for this result (around 1, see conventions), the significant result is not worth mentioning (unreliable). What you then do: just go on sampling, untill your Bayes Factor is conclusive, either for or against your hypothesis. :) (Makes things easy, doesn't it?)

Oh and your questions is very wide. So I just assumed that you are asking about the "Power" of your test. However, if your goal is parameter estimation, e.g. if you have a compuational model which supposedly captures some latent on cognitive abilities or something like this, then Michael is of course right, bigger is always better. But there it is basically also a matter of the data you have: Simpler case, just assume you want to model choice probability for a yes no response hierarchically, and in your experiment each participant gave 5 responses. This means you can model 0 out of 5, 1 out of 5 .... up to 5 out of five, given a response probability parameter, which plausibly is not very sensitive for detecting differences , e.g. in a between design were response probabilities might change from 55% to 60%, relatively independent from 'human capital'... Another issue with parameter estimation in computational modelling, is that there might be dependencies among parameter estimates (correlations as well as auto-correlations) which might reduce the "effective sample" size. Which is in fact some kind of "occupied" term here. Maybe you mean this... but this then is a matter of the model itself, and the number of iterations in the MCMC chains.

Hope this helps.

Best, René