# Why convert logodd evidence through Math.Exp

• ### Question

• Hi, Infer.net people,

From the example in the website, the probability of satisfying the constraint is computed by Math.Exp(logEvidence).

The result seems a oddRatio to me rather than a probability.

http://research.microsoft.com/en-us/um/cambridge/projects/infernet/docs/Computing%20model%20evidence%20for%20model%20selection.aspx

Anyone can help me to understand this?

Thank you very much.

Tuesday, October 21, 2014 3:12 AM

### Answers

• It is the odds ratio against an empty model, which is the same as the probability.
• Marked as answer by Friday, October 24, 2014 9:10 PM
Tuesday, October 21, 2014 10:03 AM

### All replies

• It is the odds ratio against an empty model, which is the same as the probability.
• Marked as answer by Friday, October 24, 2014 9:10 PM
Tuesday, October 21, 2014 10:03 AM
• Thank you.  Hence, Empty model means the constraint is always satisfied under Evidence=0.
Tuesday, October 21, 2014 7:34 PM