... analysis.¹

Based on the series of lectures ``Statistical methods for frontier physics'' given at the VI LNF Spring School ``Bruno Touscheck'', Frascati, Italy, May 2001.

Email: dagostini@roma1.infn.it. URL: http://www-zeus.roma1.infn.it/$\tilde{ }$agostini.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... rules²

Note that the translation of these rules into ``confidence'' is not straightforward, unless the log-likelihood (or the $\chi^2$) has a nice parabolic shape and all values where the likelihood finally concentrates are initially considered about equally likely.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... 37\%.³

This value should be compared with 39.3% obtainable in the case of a perfect 2-D Gaussian, thus indicating that the final distribution comes out naturally approximatively of that kind.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.