Between group sequential design with informed priors on both control and treatment group
Hi everyone! Thanks for the brilliant work on JASP.
I would like to use a Bayesian test with informed priors for both the control and treatment group.
The control group is supposed to have an effect centered on .10 and the treatment group is supposed to have an effect centered on .60. I would like to get a Bayes factor comparing these two models. How would I go about this in JASP?
So, rather than having a null hypothesis stating a zero effect, I would like to use the .10 distribution as a comparator for the main distribution of .60.
I appreciate any effort to help me out on this.
Thanks!
Comments
It seems to me that you want to test two different alternative hypotheses (relative to two different nulls): The hypothesis that the control condition's mean is different from 0.10, and the hypothesis that the treatment condition's mean is different from 0.60. So why not just run two separate one-sample Bayesian t tests: one for the 'control' condition and one for the 'treatment' condition?
Or if you want to know whether the control group's deviation from a mean of 0.10 is different from the treatment group's deviation from a mean of 0.60, it seems to me that you could transform the control group scores by subtracting 0.10 from each score, transform the treatment group scores by subtracting 0.60 from each score, and conduct a Bayesian two-group t test on the transformed scores.
R
Many thanks for the help. I appreciate that you took the time to clarify the possibilities for me.
Best,
D