We have performed an optimization procedure to define cuts in an unbiased
way both for the normalization mode
and for
signal separately.
We have in fact blinded the signal region for the
signal
to minimize possible experimenter bias in the analysis.
For the search mode, we derive a set of optimized cuts by maximizing the score
function
, where S is the number of
signal Monte Carlo events and B is taken from the sidebands.
This function is particularly suitable for a first observation of a decay mode
since the signal S does not appear in the denominator a normalization
factor is not needed. The optimization for the normalization mode is based on the
usual significance
, in this case the signal S, coming from Monte Carlo, needs to be
rescaled to represent the number of signal events in our data sample.
Sidebands choice for the
mode is illustrated in Fig. 1 and 2.
The optimization is performed separately for both channels and for the two
trigger paths Scenario A and LOWPT.
The variables entering the optimization are chosen with standard criteria:
low correlation, agreement of sidebands subtracted distribution with MC,
effectiveness in discriminating signal from background.
In particular the variables that we optimize in the
mode are: