How to interpret inconsistent BAMM results across different expectedNumberOfShifts values?

[Hu Ling]

Hi,

Recently, I have been using BAMM for analysis but encountered some issues. I tried setting expectedNumberOfShifts = 0.5/1/2/3/5 to check whether the posterior distributions are consistent. However, I am unsure how to interpret the current results.

When setting expectedNumberOfShifts = 0.5/1: the plotPrior results support the 0 shift, and credibleShiftSet() also indicates that the highest probability is for no shift. However, computeBayesFactors() shows that, compared to 0 shift, 1 shift has a higher probability.

When setting expectedNumberOfShifts = 2/3/5: the plotPrior results support the 1 shift, but credibleShiftSet() still indicates that the highest probability is for no shift, and computeBayesFactors() still shows that, compared to 0 shift, 1 shift has a higher probability.

I have attached both cases. These results are quite puzzling. How should I proceed next?

Anyone could help me?

Thanks

Ling

[pasca...@gmail.com]

Hi,

Before going any further, it would be important to check for some of the more straightforward issues. 

plotPrior(), credibleShiftSet() and computeBayesFactors() all have a expectedNumberOfShifts argument. Can you please confirm that you have set the value of this argument to the same value that matches what you set in the BAMM control file?

-Pascal

[Hu Ling]

Hi, Pascal

Yes.  The expectedNumberOfShifts in plotPrior(), credibleShiftSet(), and computeBayesFactors() is consistent with the value specified in my BAMM control file.

Best regards,

Ling