blob: e1233c3f4fa662563556d51eaa6cf2c2e44c41e4 [file] [log] [blame]
/**************************************************************
*
* Licensed to the Apache Software Foundation (ASF) under one
* or more contributor license agreements. See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership. The ASF licenses this file
* to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance
* with the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing,
* software distributed under the License is distributed on an
* "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
* KIND, either express or implied. See the License for the
* specific language governing permissions and limitations
* under the License.
*
*************************************************************/
// MARKER(update_precomp.py): autogen include statement, do not remove
#include "precompiled_sc.hxx"
// #include <math.h>
#include <tools/debug.hxx>
#include <rtl/logfile.hxx>
#include "interpre.hxx"
double const fHalfMachEps = 0.5 * ::std::numeric_limits<double>::epsilon();
// The idea how this group of gamma functions is calculated, is
// based on the Cephes library
// online http://www.moshier.net/#Cephes [called 2008-02]
/** You must ensure fA>0.0 && fX>0.0
valid results only if fX > fA+1.0
uses continued fraction with odd items */
double ScInterpreter::GetGammaContFraction( double fA, double fX )
{
RTL_LOGFILE_CONTEXT_AUTHOR( aLogger, "sc", "er", "ScInterpreter::GetGammaContFraction" );
double const fBigInv = ::std::numeric_limits<double>::epsilon();
double const fBig = 1.0/fBigInv;
double fCount = 0.0;
double fNum = 0.0; // dummy value
double fY = 1.0 - fA;
double fDenom = fX + 2.0-fA;
double fPk = 0.0; // dummy value
double fPkm1 = fX + 1.0;
double fPkm2 = 1.0;
double fQk = 1.0; // dummy value
double fQkm1 = fDenom * fX;
double fQkm2 = fX;
double fApprox = fPkm1/fQkm1;
bool bFinished = false;
double fR = 0.0; // dummy value
do
{
fCount = fCount +1.0;
fY = fY+ 1.0;
fNum = fY * fCount;
fDenom = fDenom +2.0;
fPk = fPkm1 * fDenom - fPkm2 * fNum;
fQk = fQkm1 * fDenom - fQkm2 * fNum;
if (fQk != 0.0)
{
fR = fPk/fQk;
bFinished = (fabs( (fApprox - fR)/fR ) <= fHalfMachEps);
fApprox = fR;
}
fPkm2 = fPkm1;
fPkm1 = fPk;
fQkm2 = fQkm1;
fQkm1 = fQk;
if (fabs(fPk) > fBig)
{
// reduce a fraction does not change the value
fPkm2 = fPkm2 * fBigInv;
fPkm1 = fPkm1 * fBigInv;
fQkm2 = fQkm2 * fBigInv;
fQkm1 = fQkm1 * fBigInv;
}
} while (!bFinished && fCount<10000);
// most iterations, if fX==fAlpha+1.0; approx sqrt(fAlpha) iterations then
if (!bFinished)
{
SetError(errNoConvergence);
}
return fApprox;
}
/** You must ensure fA>0.0 && fX>0.0
valid results only if fX <= fA+1.0
uses power series */
double ScInterpreter::GetGammaSeries( double fA, double fX )
{
RTL_LOGFILE_CONTEXT_AUTHOR( aLogger, "sc", "er", "ScInterpreter::GetGammaSeries" );
double fDenomfactor = fA;
double fSummand = 1.0/fA;
double fSum = fSummand;
int nCount=1;
do
{
fDenomfactor = fDenomfactor + 1.0;
fSummand = fSummand * fX/fDenomfactor;
fSum = fSum + fSummand;
nCount = nCount+1;
} while ( fSummand/fSum > fHalfMachEps && nCount<=10000);
// large amount of iterations will be carried out for huge fAlpha, even
// if fX <= fAlpha+1.0
if (nCount>10000)
{
SetError(errNoConvergence);
}
return fSum;
}
/** You must ensure fA>0.0 && fX>0.0) */
double ScInterpreter::GetLowRegIGamma( double fA, double fX )
{
RTL_LOGFILE_CONTEXT_AUTHOR( aLogger, "sc", "er", "ScInterpreter::GetLowRegIGamma" );
double fLnFactor = fA * log(fX) - fX - GetLogGamma(fA);
double fFactor = exp(fLnFactor); // Do we need more accuracy than exp(ln()) has?
if (fX>fA+1.0) // includes fX>1.0; 1-GetUpRegIGamma, continued fraction
return 1.0 - fFactor * GetGammaContFraction(fA,fX);
else // fX<=1.0 || fX<=fA+1.0, series
return fFactor * GetGammaSeries(fA,fX);
}
/** You must ensure fA>0.0 && fX>0.0) */
double ScInterpreter::GetUpRegIGamma( double fA, double fX )
{
RTL_LOGFILE_CONTEXT_AUTHOR( aLogger, "sc", "er", "ScInterpreter::GetUpRegIGamma" );
double fLnFactor= fA*log(fX)-fX-GetLogGamma(fA);
double fFactor = exp(fLnFactor); //Do I need more accuracy than exp(ln()) has?;
if (fX>fA+1.0) // includes fX>1.0
return fFactor * GetGammaContFraction(fA,fX);
else //fX<=1 || fX<=fA+1, 1-GetLowRegIGamma, series
return 1.0 -fFactor * GetGammaSeries(fA,fX);
}
/** Gamma distribution, probability density function.
fLambda is "scale" parameter
You must ensure fAlpha>0.0 and fLambda>0.0 */
double ScInterpreter::GetGammaDistPDF( double fX, double fAlpha, double fLambda )
{
RTL_LOGFILE_CONTEXT_AUTHOR( aLogger, "sc", "er", "ScInterpreter::GetGammaDistPDF" );
if (fX <= 0.0)
return 0.0; // see ODFF
else
{
double fXr = fX / fLambda;
// use exp(ln()) only for large arguments because of less accuracy
if (fXr > 1.0)
{
const double fLogDblMax = log( ::std::numeric_limits<double>::max());
if (log(fXr) * (fAlpha-1.0) < fLogDblMax && fAlpha < fMaxGammaArgument)
{
return pow( fXr, fAlpha-1.0) * exp(-fXr) / fLambda / GetGamma(fAlpha);
}
else
{
return exp( (fAlpha-1.0) * log(fXr) - fXr - log(fLambda) - GetLogGamma(fAlpha));
}
}
else // fXr near to zero
{
if (fAlpha<fMaxGammaArgument)
{
return pow( fXr, fAlpha-1.0) * exp(-fXr) / fLambda / GetGamma(fAlpha);
}
else
{
return pow( fXr, fAlpha-1.0) * exp(-fXr) / fLambda / exp( GetLogGamma(fAlpha));
}
}
}
}
/** Gamma distribution, cumulative distribution function.
fLambda is "scale" parameter
You must ensure fAlpha>0.0 and fLambda>0.0 */
double ScInterpreter::GetGammaDist( double fX, double fAlpha, double fLambda )
{
RTL_LOGFILE_CONTEXT_AUTHOR( aLogger, "sc", "er", "ScInterpreter::GetGammaDist" );
if (fX <= 0.0)
return 0.0;
else
return GetLowRegIGamma( fAlpha, fX / fLambda);
}