Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Supported by

Null hypothesis for one-sided Bayesian t test

Hi

As I understand it, a one-sided Bayesian t-test will produce a Bayes Factor, BF+0, which indicates the relative evidence from the data in favour of a hypothesis for a positive effect (greater than zero) versus evidence in favour of the null hypothesis. 

My question is whether that null effect is specifically either: (A) a point-null of zero; or (B) everything other than the tested hypothesis - i.e., any negative effect as well as zero effect. (I'm specifically referring to the case in JASP where a one-sample Bayesian Student t-test is run, with the Alternative Hypothesis being "> Test Value".)

Many thanks in advance!

Comments

  • The BF+0 is a comparison to the point null.

    If you want BF+- (comparison between effect postive vs negative) you can get it manually through transitivity: BF+- = BF+0 * BF0-

    EJ

  • That's fabulous - very many thanks for your help, EJ! :)

Sign In or Register to comment.