The Confidence Limits working group has released DØ Note 2775
describing the limit setting technique advocated by the Search
Limit Committee as the best available method for the construction of search
limits. This Note is available here as a
HTML document or as a
PS file.
For some information about Bayesian statistical inference, see the web page
http://d0server1.fnal.gov/users/paterno/public_html/probability
Probability Theory Web page
Other Statistical Issues
Iain Bertram has written a paper on the problems of using
an ansantz function to estimate the background when your signal is
a resonance on top of a continuum background. This paper can be found
here
Optmizing on signal / sqrt(background) is NOT the correct thing to do if either your expected signal or expected background is less than 10 events. Several people have worked on better estimators.
Here is a piece of Fortran code by Rich Partridge that will return a better estimator of the "goodness" of a cut for a given signal and background. Documentation is here.
Bruce Knuteson has a D0 note on the same subject. He also has a piece of code you can use, the note contains the location.
Jim Linnemann has a D0 note on the same subject.
Sarah C. Eno