Title: Values of the reference prior for the Poi(s+b) model from JINST 7 (2012) P01012

Type Dataset Casadei, Diego (2014): Values of the reference prior for the Poi(s+b) model from JINST 7 (2012) P01012. Zenodo. Dataset. https://zenodo.org/record/11896

Author: Casadei, Diego (FHNW and UCL) ;

Links

• Item record in Zenodo
• Digital object URL

Summary

Plain text table with the values of the reference prior π(s) for the Poi(s+b) model used in the statistical inference about counting experiments, as explained in JINST 7 (2012) P01012, doi:10.1088/1748-0221/7/01/P01012, http://arxiv.org/abs/1108.4270.  The values are useful to find approximate expressions which are quicker to compute than the original prior, as explained in http://arxiv.org/abs/1407.5893 (where this dataset is referred to).

Each line is a sequence of spaces-separated values, and the file can be considered a table.  The first line starts with two strings "shape" and "rate" which represent the titles of the corresponding columns in the data table.  They refer to the shape and rate parameters defining the background prior.  Next, N signal values starting from s=0 to s=70 are reported.  They are the values at which π(s) is computed for any subsequent line.

Starting from the second line, the format is always the same.  The first two values are the shape and rate parameters defining the background prior used to compute π(s) in this line.  Next, the N values π(s=0), ..., π(s=70) are reported.  As π(0) = 1, the third column is constant (it might be useful to debug the data reading).

As explained in http://arxiv.org/abs/1108.4270, simple functional forms may be used to fit the N points (s, π(s)).  As the shape and rate parameters from the user's application may be different from those reported in this table, the following procedure shall give a very good approximation to π(s).  In the (log(shape), log(rate)) parameters space, locate the neighboring points to the user's background parameter values (in log-log scale).  Then interpolate each of the π(s) values to obtain a set of N values (a linear interpolation in log-log scale shall be sufficient).  Finally, fit these interpolated values to find the reference prior for the user's application.

More information

• DOI: 10.5281/zenodo.11896

Subjects

• Poisson model, Counting experiments, Bayesian statistics, Reference analysis, Reference prior

Dates

• Publication date: 2014
• Issued: September 27, 2014

Notes

Rights

• https://creativecommons.org/publicdomain/zero/1.0/legalcode Creative Commons Zero v1.0 Universal
• info:eu-repo/semantics/openAccess Open Access

Format

electronic resource

Relateditems