rebinning

Rebinning is the process of changing the binning of a discrete data set or histogram by aggregating counts in adjacent bins or by creating a new grid of bin edges. It is used to adjust resolution, increase statistical significance, or to match the bin structure of different data sets or instruments.

In one dimension, rebinning typically sums the counts from the original bins that contribute to each new bin. When bin edges do not align, methods such as area-overlap rebinning distribute the original counts proportionally to the overlapping fraction, preserving the total counts. In two dimensions, rebinning can be applied to images or maps by averaging or summing pixel values within the new pixels, with area-preserving schemes used to conserve total signal.

Choosing a binning scheme involves trade-offs between resolution and noise. Rebinning can change the statistical properties of the data: for Poisson-distributed counts, uncertainties should be propagated, and excessive binning can smooth away features. Nonuniform binning requires careful handling to avoid bias.

Rebinning is common in fields such as astronomy, high-energy physics, spectroscopy, and image processing, where data are collected at different resolutions or with different instruments. Software tools implement various rebinning algorithms, including flux-conserving methods and nonuniform rebinning, and analysts should select the method that best preserves total signal and uncertainties for their analysis.