Bayes factor calculation I am a newbie to Bayesian stats...I came across an article for calculating Bayes factor by rounder etal..
In this article how am I supposed to put t value..n scale r on effect size parameters. If you could help me with a specific example it would be a great help..
I have 2 groups in my sample
group A sample size 110, mean 112.76, SD 30
group B sample size 89, mean 100.4, SD 28

What are t value ( is it by routine t test ?) and scale effect for r size in this case..and how can I calculate them
Further how do we interpret the output..accompany article is a bit inaccessible to me  because of my non mathematical ground...if anyone could explain..how can I interpret these ratios..?
 A: Yes, the t value referred to in the calculator is the classical (pooled) t statistic. The r scale factor is the scale on the prior for the effect size. See the Rouder et al. (2009) paper linked at the web calculator for details. The default in the same authors' BayesFactor package (http://bayesfactorpcl.r-forge.r-project.org/) which uses the same models as the web calculator is slightly different: r=.707. However, the change from r=1 to r=.707 will not generally change your substantive conclusions.
Regarding how to interpret the output: a Bayes factor is two things. First, it is the ratio of the probability of the data under two hypotheses. So, if we have hypothesis 1 and 2, which we can call H1 and H2, then if the Bayes factor of H1 to H2 is 10, then that means that the observed data are 10 times more probable under H1 than they are under H2. 
As it turns out, this is (by Bayes' theorem) the exact number we should multiply our a priori odds by in order to turn them into a posteriori odds. That is, the Bayes factor is the evidence, because it represents our change in relative belief between H1 and H2. So, if a priori we were in factor of H2 by a factor of 5, then the previously mentioned Bayes factor of 10 would give us a priori odds of 10 * (1/5) = 2. We are this in favor of H1 by a factor of 2, after seeing the data. The Bayes factor of 10 has shifted our relative beliefs by a factor of 10. 
The particular hypotheses outlined in the Rouder et al (2009) paper are the null hypothesis H0 (that the true effect size is 0) and an alternative hypothesis that instantiates the idea that if the effect size is not 0, then it is probably something near-ish to 0 (that is, probably not 20). The prior distribution under the alternative is a guess as to where the effect size would be if it would not be 0. The Bayes factors discussed in the paper can be interpreted as the relative evidence for each of these two hypotheses, given the observed data.
