Conclusions (1)
Current DL formula yields too many false positives
Current DL formula fairly good only if counts > 100
There are much better approximate & exact solutions
New decision rules are needed
- N + 1 rule is better, but not correct
- Altshuler & Pasternak better
- Detailed Bayesian method works if a flat prior is acceptable
- Nicholson’s (1963) and Sumerling’s & Darby’s (1981)exact (classical) solution can be implemented - probably the best