uk.ac.starlink.dpac.math

Class Edsd

java.lang.Object

uk.ac.starlink.dpac.math.Edsd

public class Edsd
extends Object

Calculates quantities related to the Exponentially Decreasing Space Density prior for estimating distances from parallaxes.

Since:

15 Mar 2018

Author:

Mark Taylor

Constructor Summary

Constructors

Constructor and Description
Edsd(double plx, double eplx, double l) Constructor.

Method Summary

Modifier and Type	Method and Description
NumericFunction	calculateCdf(double tol) Integrates the PDF to provide a numerical approximation of the cumulative density function.
double	getBestEstimation() Returns the position of the favoured mode of the PDF.
Function	getPdf() Returns the PDF corresponding to this prior.
NumericFunction	getSampledPdf(NumericFunction cdf) Returns a numerical reconstruction of the PDF based on the samples used to perform the CDF integration.
double	getUnnormalizedProbabilityAt(double r) Returns the unnormalized PDF (posterior) for this prior evaluated at a given distance.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

Edsd

public Edsd(double plx,
            double eplx,
            double l)

Constructor. The units are either max and kpc or arcsec and parsec.

Parameters:

plx - parallax

eplx - error in parallax

l - distance scale

Method Detail

getBestEstimation

public double getBestEstimation()

Returns the position of the favoured mode of the PDF. This is considered to be the best estimator of the distance.

Returns:

distance PDF mode in kpc or parsec; the lower one if there are two

getUnnormalizedProbabilityAt

public double getUnnormalizedProbabilityAt(double r)

Returns the unnormalized PDF (posterior) for this prior evaluated at a given distance.

Note this differs by a numerical factor (a function of plx, eplx and l) from the similarly named function in gaia.cu9.tools.parallax.PDF.ExpDecrVolumeDensityDEM. This nominally makes no difference, since the result of this function is declared unnormalised and will generally be scaled before further use. However, in practice this method will return a definite result for parameter ranges in which the CU9 code faces numerical overflow and returns Infinity.

Parameters:

r - distance (in kpc or parsec)

Returns:

unnormalised probability density

getPdf

public Function getPdf()

Returns the PDF corresponding to this prior. This is not normalised in the sense that the area under it is unity, but it is scaled so that the best estimate equals 1.

Returns:

sort-of normalised PDF

calculateCdf

public NumericFunction calculateCdf(double tol)

Integrates the PDF to provide a numerical approximation of the cumulative density function. The output range in distance is between r=0 and the radius above which the PDF is close to zero. The output is normalised, so its value is zero at r=0 and unity at the largest r.

Careful when interpolating this. The spline interpolation for plx=40, eplx=0.75, l=1.35 gives some nasty results at high r. Linear and quadratic interpolations are OK.

Parameters:

tol - calculation tolerance

Returns:

normalised numerical CDF samples

getSampledPdf

public NumericFunction getSampledPdf(NumericFunction cdf)

Returns a numerical reconstruction of the PDF based on the samples used to perform the CDF integration. This can be used to assess the quality of the integration, at least by eye.

Parameters:

cdf - calculated cumulative density function; this is only used to supply the sample X values

Returns:

PDF evaluated at the CDF X values