poissonregime

Poisson error bars for low count statistics, detection significances.

About

Low count statistics is not that hard. It's just not Gaussian.

This package can answer the following questions:

• Given k detected objects, what are the uncertainties on the true number of objects? (poissonregime.uncertainties_rate)
• Given k "hits" out of a sample of n tries, what is the fraction and its uncertainty? (poissonregime.uncertainties_fraction)
• What is the significance of k detections, given that I expect B background events? (poissonregime.significance)
  • what if I measure the background events from counts in a "off" region?
  • what if I have additional systematic uncertainty?
• What is the probability distribution on the event rate, given a measured background rate? (poissonregime.posterior)

You can help by testing poissonregime and reporting issues. Code contributions are welcome. See the Contributing page.

Usage

Read the full documentation at:

https://johannesbuchner.github.io/poissonregime/

Licence

MIT.

Other projects

See also:

• https://github.com/giacomov/gv_significance
• https://github.com/mahnen/gamma_limits_sensitivity