Google
 

Trailing-Edge - PDP-10 Archives - decus_20tap2_198111 - decus/20-0026/smirn.doc
There are 2 other files named smirn.doc in the archive. Click here to see a list.
SUBROUTINE SMIRN

PURPOSE
   COMPUTES VALUES OF THE LIMITING DISTRIBUTION FUNCTION FOR
   THE KOLMOGOROV-SMIRNOV STATISTIC.

USAGE
   CALL SMIRN(X,Y)

DESCRIPTION OF PARAMETERS
   X	- THE ARGUMENT OF THE SMIRN FUNCTION
   Y	- THE RESULTANT SMIRN FUNCTION VALUE

REMARKS
   Y IS SET TO ZERO IF X IS NOT GREATER THAN 0.27, AND IS SET
   TO ONE IF X IS NOT LESS THAN 3.1.  ACCURACY TESTS WERE MADE
   REFERRING TO THE TABLE GIVEN IN THE REFERENCE BELOW.
   TWO ARGUMENTS, X= 0.62, AND X = 1.87 GAVE RESULTS WHICH
   DIFFER FROM THE SMIRNOV TABLES BY 2.9 AND 1.9 IN THE 5TH
   DECIMAL PLACE.  ALL OTHER RESULTS SHOWED SMALLER ERRORS,
   AND ERROR SPECIFICATIONS ARE GIVEN IN THE ACCURACY TABLES
   IN THIS MANUAL.  IN DOUBLE PRECISION MODE, THESE SAME
   ARGUMENTS RESULTED IN DIFFERENCES FROM TABLED VALUES BY 3
   AND 2 IN THE 5TH  DECIMAL PLACE.  IT IS NOTED IN
   LINDGREN (REFERENCE BELOW) THAT FOR HIGH SIGNIFICANCE LEVELS
   (SAY, .01 AND .05) ASYMPTOTIC FORMULAS GIVE VALUES WHICH ARE
   TOO HIGH ( BY 1.5 PER CENT WHEN N = 80).  THAT IS, AT HIGH
   SIGNIFICANCE LEVELS, THE HYPOTHESIS OF NO DIFFERENCE WILL BE
   REJECTED TOO SELDOM USING ASYMPTOTIC FORMULAS.

SUBROUTINES AND FUNCTION SUBPROGRAMS REQUIRED
   NONE

METHOD
   THE METHOD IS DESCRIBED BY W. FELLER-ON THE KOLMOGOROV-
   SMIRNOV LIMIT THEOREMS FOR EMPIRICAL DISTRIBUTIONS- ANNALS
   OF MATH. STAT., 19, 1948, 177-189, BY N. SMIRNOV--TABLE
   FOR ESTIMATING THE GOODNESS OF FIT OF EMPIRICAL
   DISTRIBUTIONS- ANNALS OF MATH. STAT., 19, 1948, 279-281,
   AND GIVEN IN LINDGREN, STATISTICAL THEORY, THE MACMILLAN
   COMPANY, N. Y., 1962.