Bayesian paired t-test informed prior t-distribution
I am very new to Bayesian analysis and downloaded JASP version 0.16.1.
Interested in Bayesian paired t-test with prior t-distribution and need clarification of terms.
Please define the following
Location
Scale
How do we know these values, including degrees of freedom?
I have a database with student pre and post test results on a standardized assessment.
How may I use the information in the database as a prior?
Thanks.
Mike S.
Comments
Hi Mike,
It is a prior, which means that it reflects the relative prior plausibility of the values for effect size. You can use the default settings, but if you know more then you can change the defaults. For instance, you may know the direction of the effect and do a one-sided test. Or you may change the width (scale) and mean (location) of the distribution; default settings are the Vohs prior (normal with mean .30 and SD .15) or the Oosterwijk prior (t-distribution with mean 0.35, sd 0.102, and df of 3).
If you wish to apply the test to a given data set, you cannot use that same data set to inspire the shape of the prior (this biases the test in favor of H1).
Cheers,
E.J.