Binning

The basic assumption of the analysis is that the data points have a roughly Gaussian distribution around their statistical means. Hence the energy bins should be wide enough so that there are a ``large'' number of counts in the bins. [How large is large? My sense right now is that 5 and up is okay as long as iterations are used, but I'll do some more quantitative studies.] Class C events have their own bin, which is labeled ``zero''.

The program starts with the specified number of class A bins plus the zero'th bin for class C. If you want to fuse some of the class A bins together to get more counts in the bins, type `R' at the main menu. You will then have a choice of ``rescaling'' the bins or ``combining'' adjacent bins together. Rescaling means to make the bins spaced evenly on a log scale, with the size determined by the number of bins. The number has to be a divisor of the original number so that the bin boundaries line up. Rescaling can be used to increase the number of bins as well as reduce it.

One can also ``combine'' adjacent bins together. Rescaling will destroy the effects of a ``combine'' command. Notice that the bin boundaries always will have values that were in the original set of boundaries; one cannot reset the boundaries in a completely arbitrary way. The reason for this is that the matrix would have to be recomputed if the boundaries were changed from their original values. Note also that the class C bin is special; it cannot be subdivided or combined with others.