marginal likelihood calculation using path sampling/ stepping stone sampling approach

[Maitreya Sil]

Hi,

I am trying to estimate the prior of the shape parameter of the gamma distribution of two separate calibrations. I am trying out different combinations of the shape parameter and then wanted to perform bayes factor analysis to select the best model. I have been advised to use path sampling/ stepping stone sampling approach to estimate marginal likelihood instead of Harmonic mean estimation or AICM in Tracer. I have two questions regarding this:

1) Once the analysis is over a marginal likelihood value is printed below the the screen. A separate MLE.log file is also generated which can be viewed in Tracer. The marginal likelihood value shown there and the one printed on the screen are different. So which one is the correct value?

2) How should one go about calculating the bayes factor, since, using harmonic mean or AICM is not advised?

Any help is appreciated. 

Thank you.

Maitreya