next up previous index
Next: G103 Kolmogorov Test Up: CERNLIB Previous: G101 Inverse of

G102 Kolmogorov Distribution

Routine ID: G102
Author(s): F. James, K.S. KölbigLibrary: MATHLIB
Submitter: Submitted: 15.10.1976
Language: FortranRevised: 15.03.1993

Function subprogram PROBKL calculates the Kolmogorov distribution function

P(X) = -2 ∑j=1&inf; (-1)jexp(-2j2X2)

for real arguments X.

Structure:

FUNCTION subprogram
User Entry Name: PROBKL

Usage:

In any arithmetic expression, PROBKL(X) has the value P(X) ,

where PROBKL and X are of type REAL.

Method:

Direct evaluation or using functional relations.

Accuracy:

Approximately seven digits are correct. Results smaller than 10-40 (corresponding to X > 6.8116 ) are set to zero. Note that the above formula has a statistical meaning only for "large" N (>10) .

Notes:

  1. For an experimental distribution with N events and a maximum deviation ΔN from a hypothetical distribution, P(X) with X= ΔN N gives the confidence level for the null hypothesis.
  2. To compare two experimental distributions with M and N events, respectively, one may use X= M N/(M + N)ΔN .

G103


Janne Saarela
Mon Apr 3 15:06:23 METDST 1995