Академический Документы
Профессиональный Документы
Культура Документы
To cite this article: T. W. Anderson & D. A. Darling (1954) A Test of Goodness of Fit,
Journal of the American Statistical Association, 49:268, 765-769
Taylor & Francis makes every effort to ensure the accuracy of all the
information (the “Content”) contained in the publications on our platform.
However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness,
or suitability for any purpose of the Content. Any opinions and views
expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the
Content should not be relied upon and should be independently verified with
primary sources of information. Taylor and Francis shall not be liable for any
losses, actions, claims, proceedings, demands, costs, expenses, damages,
and other liabilities whatsoever or howsoever caused arising directly or
indirectly in connection with, in relation to or arising out of the use of the
Content.
This article may be used for research, teaching, and private study purposes.
Any substantial or systematic reproduction, redistribution, reselling, loan,
sub-licensing, systematic supply, or distribution in any form to anyone is
expressly forbidden. Terms & Conditions of access and use can be found at
http://www.tandfonline.com/page/terms-and-conditions
Downloaded by [Michigan State University] at 05:41 05 January 2015
A TEST OF GOODNESS OF FIT
T. W. ANDERSON AND D. A. DARLING·
Columbia University and University of Michigan
Some (large sample) significance points are tabulated for
a distribution-free test of goodness of fit which was introduced
earlier by the authors. The test, which uses the actual ob-
servations without grouping, is sensitive to discrepancies at the
tails of the distribution rather than near the median. An il-
lustration is given, using a numerical example used previously
Downloaded by [Michigan State University] at 05:41 05 January 2015
1. THE PROCEDURE
(1)
In practice the procedure really tests the hypothesis that the sample
has been drawn from a population with a completely specified density
function, since the cumulative distribution function is simply the integral
of the density.
The test procedure we have proposed earlier [1] is the following: Let
Xl ~ X2 ~ • • • ~ X n be the n observations in the sample in order, and let
u;=F(x;). Then compute
1 n
(2) Wn2 = - n - - L (2j - 1)[log Uj + log (1 - Un-Hl)]
n j=l
where the logarithms are the natural logarithms. If this number is too
large, the hypothesis is to be rejected.
This procedure may be used if one wishes to reject the hypothesis
whenever the true distribution differs materially from the hypothetical
and especially when it differs in the tails.
Significancepoints for W n 2 are not available for small sample sizes. The
asymptotic significance points are given below:
* Work sponsored by Office of Scientific Research, U. S. Air Force, Contract AF18(600)-442,
Project No. R-345-20-1.
The authors wish to acknowledge the assistance of Vernon Johns in the computations.
765
766 AMERICAN STATISTICAL ASSOCIATION JOURNAL, DECEMBER 1954
.10 1.933
.05 2.492
.01 3.857
2. A NUMERICAL ILLUSTRATION
the Kolmogorov statistic to test the hypothesis that the population from
which the data came was normal with mean 1 and standard deviation
1/v6. By this test he found the data were consistent with the hypothesis.
We have analyzed the same data using (2), obtaining W n 2 = 1.158 which
is well below the 10 per cent significance point, and we do not reject the
hypothesis.
The computation sheet for this calculation had the following columns:
Xi, v6(xj-1), uj=F(xj), 1-U7\-i+l, log Uj, log (l-U7\-j+l) and -[log
uj+log (l-U n-i+l)]. The operation Uj=F(Xi) is simply finding the prob-
ability to the left of v6(xj-l) according to the standard normal distri-
bution.
Another test procedure uses the Cramer-von Mises w2 criterion given by
(4)
as a weight function. This function has the effect of weighting the tails
heavily since this function is large near u=O and U= 1. It is this weight
function (6) which we treat in the present note.
Formula (2) is obtained by writing (4) as
768 AMERICAN STATISTICAL ASSOCIATION JOURNAL, DECEMBER 1954
1 [F,,(x) - F(x) ]2
f
CC
- W,,2
n = -00 F(x) [1 _ F(x)] dF(x)
rl-F(x)12
f
cc
tion of terms gives (2). The formula (2.5) in [1] cannot be used directly
here, for that formula requires that fo 1 !/I(u)du < 'Xi, which is not true
of (6).
and that the inversion of this characteristic function gave for the limiting
cumulative distribution of W,,2 the expression
-
V2 00 (-- l)ir(j -I- !)(4j +
L--·--.----·----·,--·---- e-[{4HI) .. ]/(Sz)
1) . 2 2
(8)
Z ;=0 j!
The terms of this series alternate in sign and the (j+ l)st term is less
than the jth term, j?; 1; thus the error involved in using only j terms of
this series is less than the (j+l)st term for j?;O. By using the fact that
ez/ [S(w2+ l) ] ~ez/s, one can easily verify 'that to compute the prob-
abilities correctly to four decimal places, one needs only the O-th term
for the first two significance points and the O-th and L-st terms for the
third significance point. The laborious part of the computation is the
evaluation of the integral. Let [("1j+1)1r!2vz)]w=y; then the inte-
grand isf(y)e-i Il2• The y-axis was divided into intervals according to the
integral e- i ll2 and numerical integration was performed.
A TEST OF GOODNESS OF FIT 769
The moments of the asymptotic distribution are fairly easy to obtain
from formulas given in [1]. The first two are
00 1
lim E(W,,2) E(W 2) =
"
L: - - = 1,
H j(j + 1)
00 1 2
lim Var (W,,2) = 2 I: =- (11"2 - 9) "" .57974.
n....... ;=1 j2(j + 1)2 3
The asymptotic significance points are computed to assure the prob-
abilities (significance levels) to be correct to four decimal places.
Downloaded by [Michigan State University] at 05:41 05 January 2015
REFERENCES
[1) Anderson, T. W., and Darling, D. A., "Asymptotic theory of certain 'good-
ness of fit' criteria based on stochastic processes," Annals of Mathematical
Statistics, 23 (1952), 193-212.
[2) Birnbaum, Z. W., "Numerical tabulation of the distribution of Kolmogorov's
statistic for finite sample size," Journal of the American Statistical Associa-
tion, 47 (1952),425-41.
[3) Birnbaum, Z. W., "Distribution-free tests of fit for continuous distribution
functions," Annals of Mathematical Statistics, 24 (1953), 1-8.
[4J Lehmann, E. IJ., "The power of rank tests," Annals of Mathematical Statis-
tics, 24 (1953), 23-43.