Number, no hyphen (except as a compound adjective), lower case "s"

Particle physicists describe the certainty of a result on a scale that goes up to 5 sigma. One sigma could be a random statistical fluctuation in the data, 3 sigma counts as evidence, but only a full 5-sigma result is a discovery. By definition, the probability that a 5-sigma result is wrong is less than one in a million.

1 sigma

3 sigma

5 sigma, etc. 

The certainty of the result was given as 3 sigma.

It was a 3-sigma result.