RADIO SCIENCE Journal of Research NBSjUSNC-URSI

Vol. 68D, No.9, September 1964

Statistical Inference for Rayleigh Distributions

M. M. Siddiqui 1

Contribution From Boulder Laboratories, National Bureau of Standards, Boulder, Colo.

(R eceived D ecember 6, 1963; r evised May 7, 1964)

The main infere nce proble ms related to t he Rayleigh distribution a re the es t im at iop of
its parameter a nd the t est of t he hypothesis that a given se t of observation s is from s uch a
di str ibution . It is s hown that (in case of radio signals) t he most effieient es timate of the
parameter is obtain ed using t he sample mean power. Complications may a ri se wh en data
are missin g or a re a utocorrelated. Methods are given to d eal w it h s uch complicat io ns a lso.

I. Introduction (1 ) Com])lete incoherence, whi ch mea ns that each

of A t is dist ri buted uniformly o,-er t he interval
In a preyious paper [Siddiqui, 19621 some problems (0, 27r).
of estimation and testin g of hypotheses were di s- (2) I ndependence of phase and amplitude, i. e.,
cussed in co nnection with t he Rayle ig h d ist ribution. R i and A i are indepe ndent random Yclri ables.
In the present survey article, for the sake of complete- (3) lbsence of a dominant vector, defined by the
ness, so me parts or t he previous paper will be re- condition of t he Ce ntral Limit Theorem [Cramer,
peated. Howe, -er , an atte mpt will be made to 195 1, pp . 215- 2 16, 185- ] 86]. If t he ?Lrst two con di-
conden se the material already co \rered . tions f),re satisfied , this th ird co ndi tion will be
sftti : 1ed if: (E'Z:~RDII3 (E'Z: ltD - 1I2 0 as n - HXJ .
1.1. Nota tion Thus, in particuhr, if R i have id entical di stributions
or if Ri= r, a cons tant for all i, t hen (3) is satisfied .
The function P will be llsed generi cally for any Th e cond itio n (3) will not be satis:fied if, say, a
distribution function and J) for any den sity Jun ction. domi na nt \'ector is prese nt. Under Sll Ch circum-
Th118 P(x) and P(y), in ge nernl , will not be the sam e s tances the .t\akagam i-Rice dist ribu tion applies
functions. If X is ft rftndom vftriftble EX will [Nakagami , 1940; R ice, 1944, tllld 19451. .
stand for the expected value of X find yur X for the If tho ftbo "e assu mptions hold R cos A and H sm
yariance~of X. Thus A are asy mptotically inde pendent n or mal vari ates,
eac h wi th men.1l zero and \'urian co (] /2) <T2, where
P(x) = Pr(X::; x) , EX= .L"'"" xdP(x) , <T2= ER2. The d istribution of ( H, 1) is approxi-
mately given by

and if P(x) is differentiable, P'(x)= p(x) .

and R and A a re asymptoticall y indepe ndent. We,
2 . General Properties of Rayleigh t herefore, co nsider a l"n.ndom "ariable H , whi ch has
Distributions the probability density a nd distribution function ,
respecti vely
'iVhen an incoming plane wave (electromagnetic,
sound, or some other kind) passes throug h a scat- p(r) = 2<T _ 2re _ r2 / 2 ,P (r) = l -e _ r2 / 2 ,O::;r
U U
< 00.

tering medium , which at each instant of time may

be considered as a sample from an ensemble of media This distribution is known as the Rayleigh am-
with stable
....,
statistical characteristics, the outgoing plitude distrib ution. We immediately ha "e p (O)=
wave R = (R cos A , R sin A) may be thought of as p(oo) = O, and per) has a unique mftximum at the
composed of many independent random wavelets mode ro= 2-!l2CT = 0.707CT. The qHantilefnnct ion, rep ),
...., that is, the in verse function of P er) is gi ven by
R i= (Ri cos A i, R i sin A i), i = l, 2, . . . ,n. The
problem in its most generali ty...., will impose no re- r(p) = <T[ - ln (l - p)JI j2, O::; p< l , (2.2)
striction s on t he distributions of R i . However, under
certain special circumstances, it may be reaso nable and gives t he 100 p percentile, rep ), of the distribu-
to assume the following: tion (2.1). Thus if [- In (1_ p )]112 is plotted hori-
zontally against rep) vertically we obtain a straight
I Now at Colorado State U ni versi ty, F ort Collins, Colo. line through t he origin with slope CT. The values
1005
of r(p) /u are given in table 1 column 2, in particular, We note that the variance and also other central
the median moments of Q are independent of the parameter 'Y
1' (0.5) = u(ln 2) 1/2= 0.832u. This is not unexpected sin ce the distribution of
(Zjy), hence that of In Z- 1n'Y, isfreeofparameters.
For k > - 2 The quan tile function, q(p ), of Qis easily evalu ated
without explicitly writing down t he distribution
function of Q. In fact, t he events " Q ~ q(p )" and
(2.3) "Z~ z (p)" have the same probability if q(p ) = 10
10glO z(p ) .
The mean, EE , and the variance, varR, are Thus
q(p) = 10 10g lO'Y + 10 10glO( z(p ) h ).
ER = (l j2)7l'1 !2u= 0.866u, val' R = (1-7l'j4)u 2= 0.2146u 2 •
Hence q(p) -q(0.5) is free of the parameter 'Y. This
Note that ER> r(0.5 ) > 1'0 , which indicates that the quantity, which measures the distance of1the 100
right-hand tail of the probabili ty density function is p percentile from the median, is tabu lated ~in table
much longer than the left-hand one (positi ve-skew- 1 column 4.
ness) .
T A BLE 1. P ercentiles of R , Z, and Q
Set Z = R2, y = u2. If R is identified as t he ampli-
tude of a wa\re, then Z is its power and 'Y = EZ the (I ) (2) (3) (4)
q( p ) - q(0.5 )
mean power. The probability density and dis- 100 P r(p )/" 2( p )/r
--
tribution functions of Z are , respectively, --~

0.1 0.0316 0.00100 - 28.4

I . 100 . OJOI - 18. 4
5 . 227 . 0513 - II. 3
10 . 324 .1 05 - 8.18
25 . 537 .288 - 3. 82
(2.4) 50 . 832 .693 0.00
75 1.18 1. 39 3.01
gO 1. 52 2.30 5. 21
This distribution is known as Rayleigh power distri- 95 1.73 3.00 6.36
99 2.14 4.61 8. 22
bution or exponential distribution. The density func- 99.9 2. 63 6.91 9.98
tion has its maximum at z = O and decreases mono-
tonically to zero as z increases to in:fi.nity. If z(p )
is the quantile function, then z(p) = r2(p), i.e. , A comparison of (2.4) with (2.1), or (2.5) with
(2 .2) indicates that the properties of the Rayleigh
z (p) = - 'Y In (l - p) , O~p < 1. (2.5 ) power distribution are much neater and better
suited for statistical estimation and for testing of
If - In (l - p) is plotted horizontally agai nst z(p ) statistical hypotheses. In fact, if we even start
vertically we obtain a s traight line through the with (2. 1), we find that the maximum likelihood
origin with slope 'Y. The valu es of z(p)jy are given (also the unbiased sufficient) estimator of u 2 from
in table 1 colum n 3. If k > - 1 random sample R l, . . . , EN is N - l J:,~Ri = N-l
J:,~ Z i , which is readily understood as (sample)
"mean power" rather than "mean of ampli tude
square." Similarly, the l'eparametrization , with 'Y
replacing u 2 , simplifies our estimates. From here
Thus EZ= 'Y, val' Z = 'Y2. Differentiating (2.6) j onwards, therefore, we will mainly study the Rayleigh
times with respect to k and then setting k = O, we power distribution as given in (2.4) .
obtain
3. Estimation
E [In (Zh) ]i= 1 00
(1n z)je-'dz= r U )(I ), j = O, 1, ...
3.1. Estimation From a Random Sample

Thus The only essential parameter to be estimated in a

Rayleigh power distribution is the mean 'Y. All other
E (In Z ) = In 'Y + '" (1) = In 'Y - 0.577 char acteristics of the distribution such as the distri-
bution fun ction, momemts and percentiles are func-
val' (In Z) = "" (1) = 7l' 2/6= 1.64493, tions of 'Y. Let ZI , . .. , ZN denote a complete
sample of N independent observations from the dis-
where ", (x) = r' (x) jr(x). It is custom ary to measure tribu tion. The likelihood fun ction (the joint prob-
power in decibels. Thus, writing Q= 10 loglo abili ty density of Zl, . .. , Z,v considered as a
Z = 20 10glO R , we have function of 'Y ) is given by
EQ = (10 10glOe) E(ln Z) = 4.343 (In 'Y-0.577 )

val' Q= (10 10gl oe)2 val' (In Z) =,3l.0 (2.7) This shows that J:,Z i a sufficient statistic for 'Y'
1006
Since E'Z,Zi=Ny, the unbiased 'sufficien t (also the and val' c1l 2 are possible. In fact, from (3 .2) and
maximum likelihood) estimator of I' is the sample (3.3)
mean,
(3. 1) E1 Cl /2 f (N + 1/2) l /2 1/2_ [1 fZ(N + 1/2)J
N l/2r (N) I' ,val' c - Nf2(N) 1'.
The variance of c is, of course, val' c= I'z/N, and if c'
is any other unbiased estimator of I' then Thus an unbiased estimate of ER and its variance
val' c<var c' . Since 2Z dl' is a X Z variate with 2 deg are
of freedom and Z1, . . . , ZN are independent, *_
(7rN) 1/2f (N) 1/ 2 • * _ ~( Nr2(N)
m - 2f(N + 1/2) c ,Val m - 2 f 2(N + 1/2) ~ )1"
(3.8)
is a XZ variate with 2N deg of freedom . Thus
For small N these exact expreSSlOns can be used .
X = Nch (3 .2) For large N,

has the distribution m*= 1/2 (1 + 8~) (7rc)l /2+ 0 (N- 2),
(3.3) val' m* = ·~ +O(N-Z)
16N '
from which the distribution of c is easily deri Ired by
substitution from (3.2). We note that if t>- N w hich is an improv ement over the biased estimate
m. On the other hand, let R = N - I 'Z,fR i, then
EX' = [r (N) ]-11'" xI+N - 1e-xdx = r(t + N) /r (N). ER = ER,
(3.4) ,.... 4- 7r
val' R = 4N "(.
Since c is the sufficient statistic for the family of
Rayleig h distributions { P( z) : 0::;: 1'< co}, all other Thus the r elati ve asymp totic efficiency, e(R), of 11, is
characteristics of the distribution s hould be estimated
i.n terms of c. For exampJe
e(R ) = "arm* 7r 091
varR 4 (4- 7r) .,
(3.5)

is the maximum likelihood estimator of P( z) and i. e., the sample mean 11, is a less efficient estimator
of the population mean ER than the estimator m *
based on the suffi cien t statistic c.
~(p) =- c In (l - p), o::;:p< co, (3 .6) If the m easurem ents are in decibels, that is, if our
sample consists of Qi= 10 10glO Z i, i = .l , 2, ... '. ,N,
that of z( p). In general, if a function }(I') has an then we note from (2. 7) that an estImatIOn of EQ,
unbiased estim ator, it also has a n unbiased estimator which involves In 1', is required. "Ve can directly
based on the suffi cient statistic c, and th is estimator evaluate E(In c) and var (In c) by differentiating
is the most efficient estimator of}(I'). Also let hex) (3.4) once and twice wi tIL respect to : and then
be a function which in so me neighborhood of the setting t= O. We note that
point X= I' is continuous and has con tinuous deri Ira-
tives of the first and second order. Then, if E(ln X) J= f (J) (N) /f (N) , j = l , 2, ... ,
Ih(x) l<eKX for all x2': 0 where K is some constant
independent of x, it can be shown that, for suffi- N- l
ciently large N, 1/; (N)= r' (N) / f (N) = 1/; (1)+ L, 1'-1,
r=1

(3. 7)
Thus, since In c= ln X + ln I' - ln N,
For exampJe if we wi sh to estimate the mean E (In c) = ln I' - ln N + 1/;(N) = ln I' + O (N- I )
amplitude ER = (7rl')1 IZ/2, we consider m = (7rc) 1/2/2
and from (3.7) var (In c) = var (In X) = 1/;' (N).
Thus, to order N - I, the unbiased estimator of EQ
based on cis

However, III this case, exact evalu ations of Ec 1lZ Q* = 4.343 (In c- O.577 ), val' Q* = (4.343)21/;' (N). (3.9)
1007
On the other h and, if Q=N- 1 ~Qi' EQ=EQ and confidence limits for "(. If N?,IOO, these limi ts are
approximately Nl /2c(Nl/ 2±Xa)-1.
-Q (4.3437r) 2
val' 6N' 3 .3. Estimation From Order Statistics

An estimate of"( can be made wi th reaso nably high

Thus the efficiency eN(Q) of Q relative to Q* is efficiency from a few order statistics. Let Z1, .. .,
ZN denote , as before, a sample of N independent
observations on Z and let Y 1 ::; Y~ ::; . . . ::; Y N be
their ordered values, e.g., Y1 =min (ZI' . . . , ZN),
YN = max (ZI' . . . , ZN)' The probability density
For N = I, eN(Q)= I, otherwise eN(Q) < l. Using fun ction of Y k is given by
the Euler-Maclaurin formu la
N!
p(y) (k- l ) l (N- k ) !"( exp [- (N-
~j(r) = ( kj (x)dx-I /2[j(k)-j(0)]
r-O ) 0 lc + l )y/'Y][ I -exp (_y/'Y))k-I,
+1~ [Pl) (k)-fl) (0)]- ... (3 .10) The moment generating function of Y k is

to sum the series in eN(Q) we obtain , for large

N(?10),
From <I\(u) we easily evaluate, with ai = (N-i + I) - I,

Thus the efficiency of Q is approximately 61 per- (3. 12)

cen t.
We have thus established the fact that when a Using the Euler-Maclamin formu la (3. 10) to approxi-
complete random sample is available, one should ma te these summations, I have shown elsewhere
invariably use estimates of population chara('ter- [Siddiqui, 1963] that the optimum unbiased estimator
istics using appropriate functions of the sufficient from a single order statistic corresponds to
statistic c. The use of the other estimator s men-
tioned above involves considerable loss of efficiency. k~0.7968 1 (N + 1) - 0.39841 + 1.16312 (N + 1)- 1
3.2. Confidence Limits (3. 13)

We noted that 2Nc/'Y is a x 2 variate with 2N deg where the yalue of k thus obtained is rounded off to
of freedom . To set up con:6.dence limits on 'Y with the nearest integer. For this k,
a confidence coefficient 1- a we determine from x 2 k
tables two numbers a and b, correspondin g to 2N cI = Y k /22 a i ~0.6275Yk (3. 14)
deg of freedom, such that 1

is an unbiased estimator of 'Y with efficiency

Then a::; 2Nch::; b has probability 1 - a and

(3.11 )
Comparing the efficiencies CN (CI ) and eN( Q), we
are the desired confiden ce limi ts. If h is a monotone notice that even a single order statistic, prop erly
function from (0, 00), then h(2Nc/ b) and h (2Nc /a) chosen, is more efficient than the mean of all the
will be 100 (1 - a) percen t confidence limi ts [or h ('Y) . decibel val u es for estimating the parameter of a
Thus, for example, EQ and P( z) are monoton e func- R ayleigh distribution. It may also b e noted that.
tions of ,,(, and confidence li mits for them can be the efficiency of the estimator based on the median.
easily written down. i.e., the order statistic corresponding to k~ (J /2)
If N?, 15 , x = (4Nc/'Y)1 /2- (4N- l )I /2 is approxi- (N+ l ), is only 48 percent.
mately a standard normal variate [Cramer, 1951 , The asymptotically (N----'7 oo) optimum unbiased
p. 251]. If Xa is the number such that estimator of "( from two order statistics is given b y

c2 = 0.523Ym + 0 .179Yn , m~0.639(N+ 1),

n~0.927(N+ 1 ), (3 .1 5)

then 4Nc[(4N-I)I/2 :l: xa] - 2 are 100(1 -0') percent with asymptotic efficiency e(c2) = 0.82. Sarhan,
1008
 --_._--

Greenberg, and Ogawa [1963] h ave gi ven asymptoti- Generally , the correllction function a (s) will be
cally optimum unbiased estimators of ')' fr om 1 to 15 unknown and will h ave to be estimated from the
order statistics. A linear combination of as few as sample. Let
fi ITe properly chosen order sta tistics provides an
estimator of')' whose efficiency exceeds 94 p er cent.
A big ad ITan tage of the es ti mators based on order
statistics is that only a p ar tial knowledge of the
C(s) = (T -s) - I i T
-
S
Z(t)Z (t +s)clt.
Then
sample is required. They ar e especially good when a(s) =C(s) /c~- l
low values of the sample are poorly r ecorded, not
recorded or mi ssing . L e t us say t hat the lowest 60 is a consistent estim a tor of a(s). Usually, it is
percent of the sample is missing. ' Ve can still desirable to fit to a(s) som e ma t hemati cally specified
construct the estimators CI and C2, although the function such as exp (- ,u lsl), exp (- ,us"), ,u(,u+S2)- ].
sample mean (eve n median) can not be determined. Before choosin g an approximating fun ction , howeyer ,
The optimum estimators from more than one order a more carehJ in vestigation of a(s) n ear s = O is
statistic cann ot be co nstructed if, say the highest n ecessary. For example, if we tak e exp (-,u lsl) to
8 p er ce nt or more valu es in the sample are missing. represent a(s) then, sin ce thi s f unction is Ilo t differ-
If only the highest values are missing so that we are entiable ftt s = O, we will be co mp elled to co nclude
left wit h Y 1 , . . . , Y m , m~N, the most effi cient that the process Z(t ) is n ot difrer enti abIe, a nd this
unbiased estimator of ')' fr om these order statistics is may not be a I'er y desirable sta te of ,tfLtirs .
For t he sak e of sim pli city we Lak e the u ni t of ti me
(3 .16) such t lliLt T = I . L et

and has efficiency e(cm) = m /N. In fact, 2mc mh is a i = l, 2, ... , N,

x2 variate [Epstein and Sob el 19541 with 2m deg of
freedom.. Confidence limi ts b ased o n Cm can be wh er e N is gi I'en different I"<tlues, say, 25, 50, 100,
constructed in exactly the same way as with c. etc. 'vVe calculate
3.4 . Estimation From Correlated Observations N
SN = ~ YiYi-l, and NSN'
1
In tbis section we co nsid er Z = Z(t) to b e a sta tion-
ary Rayleigh process and assum e t hat for all t a nd s It can b e shown that
E Z (t ) = ')', E Z (t) Z (t + s) = ')'2( I + a (s), (3 .17)

so that a(s) is the correlation fun ction of t he process.

N ote th ata(O) = I , a( -s) = a(s). Let Z(t) b e observed wh er e a' (- 0) is t he left-hand and a' (+ 0) is the
o ITer an in ter val of time O ~ t ~ T. From t he a nalogy rig ht-ha nd deri lratil e of a(s) at 8= 0. If a(s) is
of the random sample estim ate, we u se the sample differe nt iable at zero t hen this Ii l1li t is ;"eI"O, oth er-
mean wise n ot equal to zero. Thus Lhe beh a lTiol" of SN as
cT = T - I C
•• 0
Z (t )clt (3 .18)
N incr eases will tell us wh ether a(s) is different iable
at s= o or not. Now s llpposing a(s) is t wi ce differ-
enti able at s = o with al/( O) ~ O , we must h al'e
to estimate ')'. CT is unbiased and has I'ar iance
plim S N= O, plim NSN = - ')'2a" (0).
(3. 19) We already know, in this case, that a' (0) = O. If
exp (- ,uS2) see ms to r epresent a(s), then an estimate
We note that if we choose to take a di screte set of of ,u is obtained as m =NSN/ (2c},).
equ ally spaced observations Z(h ), Z(2h ), .. . ,
Z (Nh), then the mea n, CN, will have the I'ari a nce
4. Tests of Sta tistical Hypotheses
')'2 2')'2 N- l ( S )
val' CN= N+ N ~ I- N a(sh). (3.20) 4.1. Tests for the Distribution Function
L et Z], .. . , Z N de note indep en dent obser vations
The exact distribution of CT or CN is unknown . How- on a nonnegative I'ari ate Z. We wi s h to test the
ever , le t N' be defined by the equation statistical h ypothesis that Z h as Rayleig h power
dis tribution. W e employ the x 2 goodness-of-fit
var c= ')'2/N', i.e., N' = ,),2/var C, (3 .21 ) test for this purpose. We first calculate C= N - l ~t'Z i ;
then the numbers Xi = - C In (l-p;), p i= i /m, i = l ,
where var C is either (3 .] 9) or (3 .20) . Then 2N'c°l'h 2, .. . , m - l , where m~5. From (3.6) we see th at
or 2N' CNi'Y is approximately a x 2 type variate with Xi are lOOP i per centiles of t he Rayleig h power distri-
2N' deg of freedom. bution with ')' = C. The exp ected number of observa-
1009
tions in each of the intervals (0, XI )' (XI, x 2) , . .. , same distributions. Let C1 and C2 be the means of the
(Xm- I, ro ) is N lm . Let f1 ' . .. ,fm be the number of samples of sizes Nl and N , from Rayleigh distribu-
obsBtTed Z /s falling in these intervals. The statistic tions with means "II, and 1' 2, respec tively . Then,
under the hypothesis 1'1 = 1' 2,
m ""mf
N L.J l. 2_N
, (4.1)

is asymptotically a x2 variate with m - 2 deg of is a Fisher-Snedecor F variate with the indicated

freedom under the hypothesis t hat Z is a Rayleigh degrees of freedom. We assume that CI "2 C2; if not
variate. \,ye preassign some critical probability we simply invert the ratio and interchange the
level a ( = 0.05 or 0.01 , say) and from x2 tables de- degrees of freedom, preassign a sig ni:ficance level a
termine t h e number x~ such that Pr ( x 2 "2 x~) = a. and test for the significancE' of the calculated F by
If the obsenTed X2< x~ , we accept the hypothesis comparing it with t he upper lO Oa/2 percentage
that Z is a Raylei gh variate, otherwise we reject it. point of the F distribution. If the observations in
If the sample is truncated at one or both ends, we each sample al"e au tocorrelated, we modify the
use order statis tics to estima te I' and proceed as above degrees of freedom from N to N' accor ding to (3.2 1).
changing the fiTst and last intervals, if necessary, to If the F -test indicates acceptance of the h ypothesis
read (0, x) and (y, ro ) where X is a number slightly 1'1 = 1' 2= 1', the common mean, /" is estimated by
less than the lowest recorded observation and y is a C= N - l(NIC1 + N 2C2),N= Nl + N 2'
number slightly greater th an the highest recorded
observation. The expected number of observations 5. References
in (0, x) is N(l -e- x/ C) and in (y , ro) is N e- Y/ c• We
calculate Cram er , H. (1951), M athem a tical m ethods of stat istics
(Princeton Uni ve rs ity Press, Princeton) .
Epstein, B., a nd M. Sobel (1954) , Some t heorem s r elevant
to life t esting from an expon e nt ial distribution , Annals of
Mat h em atical Stat istics 25, 373- 38l.
where fo = obser ved , f e= expected frequency. N a ka gami, Minoru (1940) , StllCly on t he r es ulta nt l1lnplit.ude
In case of correlated observations it is necessary to of m an y vibrations whose phases a nd ampli t udes are
multiply the value of x2 thus obtained by Nfl I N random , Nippon Elec. Comm. Eng. No . 22, 69- 92.
Ri ce, S. O. (1944) , M athem aticftl ftn alysis of r andom noise,
where Nfl is the equivalent number of independent Bell System T ech. J . 23, 282- 332. (Also publish ed as
observations. Thus x~ = (Nil IN) ~ (foj!e)2
should Bell Telephone Monogn tph B 1589 and in cluded in selected
p a p ers on noi se a nd stochastic proc esses, N . W ax, cd.
Do ver Publications, N ew York, N .Y. 1954.)
be considered as an approximate x 2 variate with m - 2 Rice, S. O. (1945) , M ath ematical an alysis of r andom noise,
degrees of freedom . It is difficult to determine Nfl . Bell System T ech . J . 24, 46- 156.
However, if a (s) is the correlation function of Z (t ) Sarha n, A. E. , B . G. Greenberg, and J . Ogawa (1963) , Simpli-
process and the sample consis ts ofZ(h), . .. , Z (Nh) , fi ed estimfttes for t h e ex ponential distribut ion, Ann als of
M ath em fttical Statisti cs, 34, 102- 116.
Siddiqui , M. NL (H)62) , Some problem s connected wit h
Rayleigh distribut ions, J. R es. NBS 66D (Radio Prop.) ,
No. ~ , 167- 174.
S iddiqui, M . M . (1963) , Opt imum estim ators of the p ftr am-
cters of n egative exponen t ial distribution s fr om, one or
appears to be a reasonably good approximation t o two order st atistics, Ann als of M athe m aticftl Statistics,
Nil IN. 34 , 117- 121.

4 .2. Comparison of Two Samples

Sometimes we wish to tes t the hypothesis that

two indep endent samples, which are known to be
from Rayleigh power distributions, are from 1,he (Paper 68D9- 400)

1010