CONAN.funcs.credregionML

CONAN.funcs.credregionML#

CONAN.funcs.credregionML(posterior=None, percentile=0.6827, pdf=None, xpdf=None)#

Compute a smoothed posterior density distribution and the minimum density for a given percentile of the highest posterior density. These outputs can be used to easily compute the HPD credible regions.

Parameters:

  • posterior (1D float ndarray) – A posterior distribution.

  • percentile (Float) – The percentile (actually the fraction) of the credible region. A value in the range: (0, 1).

  • pdf (1D float ndarray) – A smoothed-interpolated PDF of the posterior distribution.

  • xpdf (1D float ndarray) – The X location of the pdf values.

Returns:

  • pdf (1D float ndarray) – A smoothed-interpolated PDF of the posterior distribution.

  • xpdf (1D float ndarray) – The X location of the pdf values.

  • HPDmin (Float) – The minimum density in the percentile-HPD region.

Example

>>> import numpy as np
>>> npoints = 100000
>>> posterior = np.random.normal(0, 1.0, npoints)
>>> pdf, xpdf, HPDmin = credregion(posterior)
>>> # 68% HPD credible-region boundaries (somewhere close to +/-1.0):
>>> print(np.amin(xpdf[pdf>HPDmin]), np.amax(xpdf[pdf>HPDmin]))
>>> # Re-compute HPD for the 95% (withour recomputing the PDF):
>>> pdf, xpdf, HPDmin = credregion(pdf=pdf, xpdf=xpdf, percentile=0.9545)
>>> print(np.amin(xpdf[pdf>HPDmin]), np.amax(xpdf[pdf>HPDmin]))
Contents