You are on page 1of 3

The Utilization of a Known Coefficient of Variation in the Estimation Procedure Author(s): Donald T.

Searls Source: Journal of the American Statistical Association, Vol. 59, No. 308 (Dec., 1964), pp. 12251226 Published by: American Statistical Association Stable URL: http://www.jstor.org/stable/2282639 . Accessed: 11/09/2013 17:14
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

American Statistical Association is collaborating with JSTOR to digitize, preserve and extend access to Journal of the American Statistical Association.

http://www.jstor.org

This content downloaded from 111.68.96.57 on Wed, 11 Sep 2013 17:14:20 PM All use subject to JSTOR Terms and Conditions

OF THE UTILIZATION OF A KNOWN COEFFICIENT VARIATION IN THE ESTIMATION PROCEDURE Inc. Research Westat Analysts, informaa priori utilizes which 9', of,uis developed An estimator, of variation. coefficient the value of the population tionconcerning mean, thesample than error squared mean to havea smaller y is shown and ofvariation coefficients various for efficiencies y.A tableofrelative sizesis presented. sample
DONALD T. SEARLS

disprior or assumed known Frequently, information. a priori ofutilizing paperHenare used.In a recent to be estimated oftheparameters tributions information. oftheuse ofa priori an excellent example dricks[1] presented variofsampling theeffects to reduce have beenproposed techniques Other by method proposed distance these is theminimum Among in estimates. ations of on truncation methods based has author studied The [2] Wolfowitz [3]. observations. extreme ofthemean,u,ofan unknown withtheestimation Thispaperis concerned estimate, unbiased thestandard information Usingonlysample population. g, a smaller mean-squared with an estimator As willbe shown, can be calculated. information. a priori byutilizing error (MSE) can be obtained availableto is sometimes information simpletypeof a priori A somewhat longassociThrough fields. thosein thebiological particularly experimenters, disposalquite material theymayhave at their ationwiththeirexperimental whichtheywill of variation the coefficients concerning accurateinformation estimate used to plan experiments, is frequently This information encounter. samplesizes,etc. in the of variation (v) can be utilized the coefficient in which The manner error or mean-squared is madeclearwhenthe accuracy estimation procedure Froma random is considered. average, of 4, a truepopulation ofan estimator where
sample of n observationsYi, Y2,...
, y', can be constructed y,,an estimator,

are based on theprinciple problem to theestimation approaches AYESIAN

- E yi, y' W
j11

(1) is

E(y'-,)2, error, so thatthemean-squared and W is chosen In thiscase,

a minimum.

MSE(y') =

nW2o-2 + ;12(l

nW) 2(2)

to W gives, Differentiating (2) withrespect


- [MSE(y')] dW
-

2n[W2

-- 2(l-nW)],

(3)

1225

This content downloaded from 111.68.96.57 on Wed, 11 Sep 2013 17:14:20 PM All use subject to JSTOR Terms and Conditions

1226 and

AMERICAN STATISTICAL ASSOCIATION JOURNAL, DECEMBER 1964

lI2 [MSE(y')] = 2n[2 + n,u]

d2

(4)

value forW is the required is alwayspositive Sincethe secondderivative gives This theequation. found by equating(3) to zeroand solving

W = 1/(n+ v2)
y'-[1/(n + v1)] j=1

(5)
n E yj

(6)

and
MSE(y')
-2/(n

+ v2).

(7)

ofthesample lessthanthe.variance From(7) it is clearthatMSE(y') is always y. average, y, is: to the sampleaverage, withrespect efficiency The relative R.E. = MSE(y)/MSE(y')
= (n + V2)/n.

(8)

in valuesofn and v arepresented fordifferent (%) oftJ' Relativeefficiencies Table 1.


TABLE 1. RELATIVE EFFICIENCIES (%) OF 9' FOR SAMPLES FROM DISTRIBUTIONS WITH SELECTED COEFFICIENTS OF VARIATION Sample Size
V

5 180.0 280.0 420.0 600.0


120.0

10 140.0 190.0 260.0 350.0


110.0

50 108.0 118.0 132.0 150.0


102.0

100 104.0 109.0 116.0 125.0


101.0

500 100.8 101.8 103.2 105.0


100.2

2 3 4 5

in practice, sizes.Frequently for smallsample The largest gainsareobtained suchsamplesizesmaybe all that are expensive, whenobservations however, are available.
REFERENCES

will fallinto a that an observation [11 Hendricks,W. A., "Estimationof the probability 59 (1964), 225-32. Statistical Association, class," JournaloftheAmerican specified [2] Searls, D. T., "On the large observationproblem,"Ph.D. thesis, North Carolina State College,1963. distancemethod,"Annals of Mathematical Statistics, [3] Wolfowitz, J., 'The minimum 28 (1957), 75-88.

This content downloaded from 111.68.96.57 on Wed, 11 Sep 2013 17:14:20 PM All use subject to JSTOR Terms and Conditions

You might also like