baseCode.math
Class SpecFunc

java.lang.Object
  baseCode.math.SpecFunc

public class SpecFunc

extends java.lang.Object

Assorted special functions, primarily concerning probability distributions. For cumBinomial use cern.jet.stat.Probability.binomial.

Mostly ported from the R source tree (dhyper.c etc.), much due to Catherine Loader.

Version:

$Id: SpecFunc.java,v 1.7 2004/12/24 23:16:09 pavlidis Exp $

Author:

Paul Pavlidis

See Also:

cern.jet.stat.gamma , cern.jet.math.arithmetic

Copyright (c) 2004 Columbia University

Constructor Summary

SpecFunc()

Method Summary

static double	dbinom(double x, double n, double p) See dbinom_raw.
static double	dhyper(int x, int r, int b, int n) Ported from R (Catherine Loader)
static double	phyper(int x, int NR, int NB, int n, boolean lowerTail) Ported from R phyper.c

Methods inherited from class java.lang.Object

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

Constructor Detail

SpecFunc

public SpecFunc()

Method Detail

phyper

public static double phyper(int x,
                            int NR,
                            int NB,
                            int n,
                            boolean lowerTail)

Ported from R phyper.c

Sample of n balls from NR red and NB black ones; x are red

Parameters:

x - - number of reds retrieved == successes

NR - - number of reds in the urn. == positives

NB - - number of blacks in the urn == negatives

n - - the total number of objects drawn == successes + failures

lowerTail -

Returns:

cumulative hypergeometric distribution.

dhyper

public static double dhyper(int x,
                            int r,
                            int b,
                            int n)

Ported from R (Catherine Loader)

DESCRIPTION

Given a sequence of r successes and b failures, we sample n (\le b+r) items without replacement. The hypergeometric probability is the probability of x successes:

 
  
   
    
     
      
       
        
         
          
                     choose(r, x) * choose(b, n-x)
           p(x; r,b,n) =  -----------------------------  =
                        choose(r+b, n)
          
                    dbinom(x,r,p) * dbinom(n-x,b,p)
                  = --------------------------------
                        dbinom(n,r+b,p)
          
          
         
        
       
      
     
    
   
  
 

for any p. For numerical stability, we take p=n/(r+b); with this choice, the denominator is not exponentially small.

dbinom

public static double dbinom(double x,
                            double n,
                            double p)

See dbinom_raw.

Parameters:

x - Number of successes

n - Number of trials

p - Probability of success

Returns: