Fitting with finite Monte Carlo statistics

Most of the following text is taken from [17] - see this publication for more discussion and details of the algorithm, and examples.

Analysis of results from HEP experiments often involves estimation of the composition of a sample of data, based on Monte Carlo simulations of the various sources. Data values (generally of more than one dimension) are binned, and because the numbers of data points in many bins are small, a $\chi 2$

minimisation is inappropriate, so a maximum likelihood technique using Poisson statistics is often used. This package incorporates the fact that the Monte Carlo statistics used are finite and thus subject to statistical fluctuations.

• The Problem.
• Methodology.
• The Solution.
• Other points concerning the solution
• Weighted Events
• HBOOK routines
• Example of fits