by
J.A. Battjes
Dept. of Civil Engineering Delft University of Technology
The Netherlands 1969
SUMMARY
There is some evidence that the heights and the periods squared of wind waves each are Rayleigh-distributed. If this is the case, it may be surmised that their joint distribution is also of the Rayleigh type. The bivariate Rayleigh distribution, which is known in the field of statistical communications theory, may thus be applicable to wind wave problems. However, 1n the field of coastal and ocean engineering, this distribution does not seem to be known. In this note attention is dra~m to its existence, and its ma1n properties are listed.
1. INTRODUCTION
2. THE BIVARIATE RAYLEIGH DISTRIBUTION ... .
2. 1. Backgr01.md
2.2. The probability density function ... ... .
2.3. The cumulative probability function ... ... .
2.4. The marginal probability density functions ... .
2.5. The moments ... ... . 2.6. The correlation coefficient ... ... .
2.7. Case of zero correlation
2.8. Case of 100% correlation • • • • • • • • • • • • • • 0' • • • • • • • • • • 2.9. Regression lines .. . . . .. .... . LIST OF REFERENCES paee 3 3 3 4 4 6 8 8 9 11
1. INTRODUCTION
In the fields of harbour-, coastal- and ocean-engineering, knowledge
of the statistical distributions of heights and periods of wind waves is
often of great importance. Many efforts, both theoretical and experimental, have been made to determine such distributions.
Empirical wave height distributions were first given by Putz
(1952).
At about the same time, Longuet-Higgins
(1952)
pointed out that theoreticallythe Rayleigh distribution should be applicable to the heights of wind waves
with a narrow spectrum, provided the waves be not too steep. It has since
been established empirically that the Rayleigh distribution:-applies to wind
wave heights with a fair degree of accuracy, even if the waves are steep and
do not have a narrow spectrum. In the latter case, the wave height may be
defined as the difference of the maximum (crest) height and minimum (trough)
height between two successive zero up- or downcrossings.
The analytical determ~nation of the distribution of wave periods (intervals
between successive zero up- or down crossings) is difficult. So far only
approximate solutions have been obtained tonguet-Higgins,
1962).
Empiricaldistribution -functions have been first proposed by Putz
(1952).
Bretschneider(1959)
noted that the squares of the periods of wind waves had a Rayleighdistribution with about the same ~egree of accuracy as the wave heights.
This conclusion seems to have also been reached by Russian authors (Bretschneider,
1965).
The distributions mentioned so far pertain to the heights (H) and periods
(T) separately. The problem of the joint distribution has not been solved by
theoretical means. Bretschneider
(1959)
approached this problem empirically.. . . . . 2
Hav~ng observed that the marg~nal d~str~but~ons of Hand T are both of the
Rayleigh type, he assumed that this would also be the case for the joint
distribution;; However, he was unable to e,laborate . - this assum:r,tion because
the bivariate Rayleigh distribution was unknown to him. To quote:
"The joint distribution of wave heights and lengths (or wave heights and
pe-riods) in general is difficult to describe completely for all conditions of
correlation, ,' .. The fact that both marginal distributions .. are 9f the same type ~s
of some help. The bivariate asymptotic problem of joint distribution for the
It was unknovm to Bretschneider that this problem, which also ar1ses 1n the
field of communications theory, had already been solved by Uhlenbeck
(1943
)
and Rice
(1944, 1945).
The fact that the theoretical bivariate Rayleigh distribution is known
seems to have remained unnoticed in the fields of coastal- and ocean-engineering.
Apparently, there is a communications gap bet.Teen researchers dealing with wind waves and those working in the field of statistical communications theory.
It is the purpose of this note to contribute to bridging that gap .. One way
of doing this would be to: merely mention a few references. However, it 1S
deemed more useful to reproduce the main results, because not all of these are
available in any single pUblication known to the author. Also, some corrections
and additions will be
made~)
It is emphasized that this note in what follm'Ts deals solely with the
the0retical bivariate Rayleigh distribution. At this time no theoretical or
empirical evidence is offered to prove or to disprove its possible
applica-bility to wind waves.
This assumption cannot mathematically be justified. In fact, Frechet
(1951, 1956
)
has shown that corresponding to each given pair of marginal distribution functions, an infinite number of bivariate distribution functions can exist.jY
Some IDlnor corrections appear in paragraph 2.5. The additions consist2. THE BIVARIATE RAYLEIGH DISTRIBUTION
2. 1. Background
The envelope of a narro\~band random noise signal has a Rayleigh distri
-bution. Similarly, the values
X
(r) andY=
X
(t+r)
of such an envelope attwo different times separated by a certain time lag ~ have a joint Rayleigh
distribution. The derivation of this distribution seems to have been given
for the first time by Uhlenbeck
(1943)
and again, independently, by Rice(1944
-
1945).
The method employed is essentially similar to the one used 1nderiving the one-dimensional Rayleigh distribution. The derivation "Till
not be reproduced here.
2.2. The probability density function (p.d.f.)
The joint p. d. f.
1(Sc
,~
)
of two stochastic variablesX
andY
isdefined by
)( and )fare said to have a j0int Rayleigh distribution if their p.d.f. is
. given by
(2)
in which a bar denotes average values and
l'
o is the modified Besselfunction of the first kind of order zero (Uhlenbeck,
1943;
Ri;e,1944-'45;
Nakagami,
1964).
The parameter Afis defined byin which C;;'(w) is the power spectral density of the narrow-band random
The case
£/
occurs when 7=0 , i.e. whenX::: Y ,
or when there is. 100% correlation between
X
andy.
For very large c, -(~ 0 , whileX
and)I
approach stochastic independence and)therefore, become uncorrelated. Although in these special cases ~equals the coefficient of correlation betweenX
andy,
this is not generally the case. The general relation-ship will be dealt with in section 2.6.
In the following, the definition of
~
as given by Eq.(3) will not be needed; ~ will merely function as a parameter which fulfills the inequalityK/~
For variables which have been normalized so as to make their average value equal, to one, the joint Rayleigh probability densitY'-becomes
(
4)
o
This is the form which will be used subsequently.
2.3. The cumulative probability function
The author has not succeeded in expressing the cumulative probability.
function, or the distribution function, in a finite number of known functions. It appears to be necessary to compute it numerically if and when the need arises.
2.4. The marginal probability density functions
Th~
marginal p.d.f. ofX
is defined by aoIC~)
= /-f{x,,~)
tiJ--Co
Substitution of Eq.
(4)
gives rise to an integral which may be evaluated by using the follow-ing series expression forID
(Abramowitz and Stegun,1965) :
:zj /;
I).:t
and by integrating term~se. The result is, as expected, the one -dimensional Rayleigh p.d.f. = 0
lor
.::t:<o Similarly> = 0 2.5. The momentsThe moment matrix U is defined by
r~?t
(8 )
Substituting Eqs.
(4)
and(6)
and integrating term~se, it can be shown thatApart from a constant factor, the infinite series ~ the right hand member is the same as the series expression for a bypergeometric function (Magnus
and Oberhettinger, 1943), so that Eq.(10) may be written as
(Middleton, 1960), or, after performing a Imovm transformation of the hypergeometric function:
(12)
(The equat ions 9.22 in Middleton (1960), corre sponding to Eqs. (10) an.d (11)
By definition resp. normali.ation, the zeroth- and first moments are equal to one. The second moments are given by
(
13)
and
The last expression is equivalent to
in vThich
!l(~)
andE
(-I)
are the complete elliptic integrals of the first and second kind of modulus1f.
Eq.(15)
is due to Uhlenbeck(1943)
.
It may be proved by using series expressions of the various transcendental functions in the right hand members of Eqs. (14) and (15), or of Eq'., (10 ) with 4 n=
'11_ I2.6.
The correlation coefficientThe coefficient of correlation between
X
andY
is defined byXy
xy
(16)
Substitution of
%.:0'
y=/
and of Eqs.(13)
and(15)
g1ves'/ '-- .J!"
'f
A graph of
;0
~
.
~
is shown in figure 1 by the full line.E
(f)
and!La)
are even functions of-:t.
It follows that.p
1S also aneven function of
t.
Two special cases occur when -(=
0 and.f
=
1:E(o)=
k(o)= :.
J --40.;0=6
Y
-;(=0
IL
(-£)
- ? 00 Q4 - - ( -+ J)--t,d {I-
-(~
kf-t)
~
0.> --<10.F
=1i
--'=1 . (19)This confirms the remarks made in section 2.2 following the definition of
~
•Utilizing series representations for ~ and }('in Eq. (17), or
for the gamma-functions in Eq. (10) with -?n- = /?t. = / , the following series
representation for
;0
may be obtained:;0=
~ Cf~
+?f
QO/.
3.
~
. . .,
C
Y·
-.J)}
Z.~
IZ'
V
J
L
~
rJ=:l.-t;.
b'g.
.
..
(2f) (20) orp
+
-
--
--
J
in which CL_ 7T (22 )It should be noted that
;0
is a non-negative function of ~.Thus, two stochastic variables with a joint Rayleigh distribution cannot be negatively correlated. Bretschneider (1959) already arrived at a similar, though more restricted, result. He proved that two stochastic variables could not be jointly Rayleigh distributed if
? - / .
In order to obtain an explicit e.xpression for ;( as a tunction of
.? '
the series given by Eq. (21) has been inverted, y,ith the result
/
(
P).:<....
/
(
'p)3
iZ
a... - /Zg ~ (23)The two series given by Eqs. (21) and (23) are rapidly converglng for
1..£ I
~ 1:,I
f J ~ I • Graphs off
1Y.).i
,
based on the first three termsof the series·, are given in figure 1. Only for values o:f';o and -,( close
to 1 is there a visible de,nation from ,the exact relationship. The truncation
error after 3 terms is less than O. 1 % for all
I-{
I
<
0/
and less thant
,till
1+t 1~1·~..1:ZP-~I'I+ ~ ~ ~-. i~ Uitl"n'!
w RP F..:.-ll ~':.l..:.. :"::1 L i.: 'r:· I~! ~: ~,~ ':..2 ~I,l 1,1~rr I~ 11.,1', II ~
!'J
I ['II f -I'i'"
~I'li
11: . ' i:~:rm~ ~!II' ~- I 'I'li'"
i1tl-l1jl- 11[,r j" C" -, , Lrr "1'1 ";.1't'"
"
I'rlf, I ,L I ;,;~ 'U, 1f!1 .p 'I fIli ·'Li ',Ii ~'I' : I ("lit-iL ," I '" , " !,~ r·i :", ' " II,' '1tf I~- . . + tiJ~~ r -,.-r tr-ll;ti: 1 8 ~ 4+'''I'--t:-I~':'k'';''LJ:::;r!t';;~~''''''':~~·l'.-:·~~:·;n.! .. ·t.
1 w.1 ri!~ ~'I' ,r,1 , , ' j If,1 ill r ,T111"1 '~ I : j : ' "ji~jP 'it!" i
lit
l.00
1 J tIT.II~7~1'
Ti
I '1 1.'1 f: ~ ~T :ll 7·' iljl " '_I; I;:; [i"r;' ler;.!
I
fl.llltr:~< 'rtf 'ti1. I, LHlj
LF
1lf J.L' .Ia
~;1
f
lft
11[1 ~l:ti 111. 1.1, Jlli dll~ rj;r ~il .rH ..:.rlt ~h Il.h rPl L: 'U1 HI! ~-:~IW' j !~:1;11'
,
1:
III, ;111 q~, it:! ,f II ~r!}1 ILlt! f ~.~ I"H+ 1.
1!fr' if -,I~H u:;! ~Il' ,,~ rrlll G r':
ii
I~t ;~llt r~:f H~' iii:r1 fl'fl' )11' ~IU" r ~Il 'if ~;~l ,-;'1' 'I'ill~! "~~I'l' ~tt- ""
OO++:J1m"f>
.
'
Lm-·
1r.\ +n,
h ,"
,- ":::
' j j " , ;;.1.,1 , . . • , "n' " ~i'
I
I, r -, I ;:f'I!
"'f III' 11'[ II ~:til 'fill
it[1
rfr i'
'iH
"!) tit:: til'
I
'IT! n.r·pc·
v"
rtl
p
-:-r
L'I
1'0' r'f IITt17
l'" ~';:iif'l,It', tI,' I,
;B~l;+
l.
hi
llHI,1
:~!~~tH
~lt:l~l J
U
"J@1
!.:-
~
1
}ritJIDi
d
&;
~
~L'E;.~R+#tt~J ~tllit'+' flfr~1;':'11j
ltH:iI'~~iE
I
;
1 H II~I III'1
~ II'c" Ir:-Ui
-,
'Jml!1 Hi:
liU: I ill ",I fm ril! Itt ifT'ti u ' "Ll'" ,..1:lr- h ,'I' Elf "~,I " 1" ti:j[t,t~ lh~ ',il
f'
I J:~,ltJ ~;iil.U'r;1
WI
I,dtj.
~b '~. If; ;'~1i.lt;
1t:l!.1:14 ±~i.~; 11 ~rrJ *1 T till 11: i i ,tij
JI..f
il':IP:
F!F!!I
Il:1l"'!
1:!itl!
I qll 'I 1'1, l i l t f i! !'1 ,it! ml ~ 1 [I; 1J ,+¥~ 'Il ~'iP.i...
T:. 'S:I !F.! :,';19tf
i'l! t!tl ' ...rA+f.
:~~ :J.:r~.;r~1 It I"I{ H+f r;:'l:r:~
m
jji '
!;p
~r , :lll '~f Ii,; In· JT It,' '" !r:4r;
!tn-II "ll· I~ fUr. I.' lill "I ,!,1 HI11~!, ,:t ~ ~ljl ,llb:=t: '~
II
[[
,
[HI
k!l nu
~~
:lii
H
L~
W!lll
'i:1
,r!
II
j~h"
fili
ti '!!'
j
lr
ur
:~ ffji
1m
iHi
IT:!
Itt
I
i
'11
j:~
l
lE
iiit
!IE
Ue
i,
J
;I,
,FI
J~
lil!
'[fF
-rti=t1
'
Q
>1~
lUI.
~
I
11
l!.~1
'18
I~
~li
~
ilifm
ill
i
r~l tm
~
fi~li~!!il:i
lit
~;
ll'
~~
U
tf
~
i
'
WW
m'r
~1
1
m
lUI
pl;
I
:r!
-tI,ll
ltF
,llf
f~1
Jt
j
~H!
~f
lr
i~
~
HJ
:
;1 im
~
till
mt
~
F.1~
f~
~
fIT~gffif~f~.~
r!rt
I
Ifi
I
fffi
~
flt~i
'
~~
l
flit::i
U
~,
~1
'
~1 ~
il
Ii .:1~
If1
R.
~
iIi,
hu
.
fr·:r
l'U
lIIH
~f1J ~f
ii
~
~
i~
fdl
t:~:~'
1.=rm
m,
lli
~
ltJ1f
~r[
.~
I
I
m
fttt
Em
~
t
·
~m!
r
'Ii . f.H11~lt
~
~~;.t
fl41.B
~
I
ili
Ili'l H-E 1l;-f~'I!I tl
l
ttt!:H.l~
r I.\,~
tltl, !f .
r+u '!~f ~1 fElt f1-i f . " - -l{1]' i F~~I-! I IIi I l:l£. +, m:.s I ,I H-Ittr+ JH , 1'
8
~ !W~h
m
.~ "!
1~~
I
~
~
~
'
m ~
~
~r
t
In
~
L
M" I~#H
'
~i
~
'
l
j"'i
:~
It
:t1
~
!
I
q
~f
f
i
t
tr·
~Hm-urJ1l
~.
1!'),fl'~11 '~:i{fPw
:W
fhH
+11-
'
",
1~j:t ,~~. ,fj: i:r st, IF :t\~ !~'h
,~I. .,1m '
if
lint
~~H~
!ai
.!
~l
1m
1
'
~l~r
&
t~jttlJ~i
M
t
J
j
lli
.:t
1
i±
J~'
I:f1m
,!t
it'
t-rmi;; b
j
~ ~
I'
~ .~}
:f ~
,
;
~
1t
~ ~
~
~
i
.I,
:
t1''!
~
~' f
~
~
1~
~~t~~
~
'
1;'
r~t- ~
1·
)
11t
iff
:t
.
~!I
I
I-R'H
t
tl'i!
J.li {:I.
~tL ~1i
I
tt
,. "
l:l'!:!.dl~'C;lHtl-'b.;.;f '1:t~ftl'l-h";C:-I-tt-~l~lU:~:i~~ ~j of JIll' .I
fffi
HI ':b! +
R-41
mf
1,t8:mt
-[
n
ffil
WJ
;i~llW
j]I~
1111wilit
mt
hl!hUH
tlli:
~H-
E
iIF
±t~
I
,
~
cf++1:III+lI,t:t1:t-1t11n
~
tttli:H!'
.
I
.
ti
m
W
tttt
'
t-
~~~~
~
fi
~-f±•
-
. 11
.
~Hifftf
l
~
f:
i:j: IIi'rJ#~
,
:lttlitfFl'ffFfFfm "., TITli"
!n tIi
"t1+,'T11J-tii~l. i';il -l, " . I~> -:r!-11, fl)~1
, 11 """j+J"~ri'-'1' !"iI-i-'ffi-'';'r:1ri'I'~ ~~t~ 1;~~,I;?;IIIII'I!!J+J+l:i:ilij:J:il1!I'=l+IIIIi1ij:1I;:W1l1Iil+Httttttt:::t:: :tj-l- II: 1;4 ~i ': I.j.f.!. 1:} j 1t.11 d ,. "'+'hl tI, H: ,I" 1,,,111 itill dn,rIf
.
if
:
f
~ t:tttt q~ ~trr
n'
:
r,
-
h
-
i
r
1+
it
~ ~
i . . r~I
~~'
I
-at1
f
i:f!Wffil
~
ll
1-/
I:$:~
Li
-il
r
~N.
-1= . f± ~ r _ :r i ~Lli, jf
I
ti~
I~t
+
.
,
,
,
..
,
r.+~
tJ-fj:1f =t:ffi 2t. - . ' I 1 Ii i, I ·~u
n
i: .t::: ~ Ij t-t~ 1-1' t i~T:IP. i'='J.+f . ,n·r T r ~W..l"",,3,+l-l~ir I~ -1= w in· Ti-· iioH ~r·li hf.tl~ ;-njrl
lif~rf ~ll '
til:
~I~r+r
d
'
~,rtf
'
.
ff
F
ijr
It . m~ ~..
~~OOf
:t ~ -l-1tJ: :['m
IT r '! ~·fit tfOJ ' ~tt
=!J ' fE(:HfI
,
' I+I-H I tw Iiltllm j:j: " !':tt rr I r-= + ~ tTl-J-t· i:t Itt;:i: fitI·H-tt fi#lAA ITt!:
ag·
ttg.t [g]:. J..:".:-'- . .!:j:I 1: ret: !ml"':':c~ ~ f?rl=.
h'tt!t.h .
,
~,'
If¥~ ~~~!:f$ ~~.: ~. tri ffi'm
l
~,
'.4-
~..
~ -r~~.li ~m
-..
Wti
- . . - - . 1 : . - I... . . ' ' ; ' .;~. q::; _ . . r Ii I - ~ : f
~r
1l1~
Wltal
:.
,It~
1- ·~1m=;
I
®~F1 ~
fUii:fli:1~~!:+;~ffflr.£~:##~;:c~~g-!,-tM
l
mJlftt::
rHtl-l:+; ' flF ~t:;.ti:;' ~~ 11-I=!+1-r 1T ::Qt:lur ITP' :1i ,- ~I ""
U+F-l+ ~r t ~~ ~ 1Jl'~ L ITIT+-f m q r~ , '-h~n!: ~ rt I f-r J I -I
1'j:
J:t
.n
~
il
l
ii
t
H
l±!:!:
lt
lIflJffillli
W:fl'l: '~
~
jj:
rri++
rm
'
F~
Ji 'tt!=: :~ i±~. ~ ~i:F::""
rrttfi1.l! -mt:.l:j::j.f.Itl:i:t:ft+...I+J±j:j::ji :Jj 1lilJ: ±b: .. tE.Jh;: ~ I± . ~ L . . ,y , ~ =tr''''''''+1+ - , -f.:t:lJ+ r I j ~ _ ~ ;+l:tl:tj:j: '.
W
ltW
1
r
l
rrt ~~Ir-~': h-'iIT
OO
'
:..t j-lttl . ' :r:t-1tH-t - ttl~ '. '-lili ,.a
_
[
lit
r:- . >.1 .J:',l
!ll
g~,. 1~ f:- ' , ' . . ~:-1+4 " I If L , liiq :J:I I1IL r ·rr H=Plrc IW-- " h1+j. ~li'-ttj:j :rl:1iL ~ \±I:I t
8tl1~ , l1g B. tIE IT ~. -t . li.q ~
.l
fili
l=!:
-
,
IIrn
ij
~ '1:$: I± j:j, ~8
H
ut.tm
,
1:l:tl:l:\:l:4i
mt:::I -~
H-I _~
j:j:j:j: r2.7. Case of zero correlation
The stochastic variables
X
and Yarecoefficient of correlation
;0,
given by Eq.Substitution of this value in Eq.
(4)
yieldsi{
)
_
rrz
_![(x.:<-r(j<-]::t.J/ -
fjX:f
-e
'<
= 0 o~
said to be uncorrelated if the
(16), is zero. If
?
=
0,(24)
which is equal to the product of the marginal p.d. functions as given by
Eqs.
(7)
and(8).
Thus, two' stochastic variables with a joint Rayleigh distri-bution are stochasticallylltlependent if they are uncorrelated.
2.8. Case of 100% correlation
If 100% correlation occurs then the two variables
X
and Yare linearly .dependent. Moreover,
X
and'I
are identically distributed. Therefore,X=
Y
iff=
1. The two-dimensional p.d.f. actually degenerates to a one-dimensional one.The joint probability density must then be zero for all
..c:/:t-
and be infinitefor .;c~o This may be shown formally by investigating the behavior of
./'rz,j-)
as~~/. To (this end, the following asymptotic expression for
.z;,
(i)
(AbramOi-litz andStegun, 1965) is substituted in Eq. (4):
After some algebraic manipulations, the result can be written as follOvTs:
(~6)
in which
. .<.
u = _
Tr
The expression J.n the brackets is formally equal to a Gaussian p.d.f. wit'h
indep~
variable (x-t) .. zero mean and mean' square deviationcr-. As -;f!approaches 1,0- approaches O. The Gaussian p.d.f. then a~proaches Dirac's unit
as a definition of
d
.
Thus,(28)
which may also be v~itten as
or as
(30)
in which
~(x)
and~(1)
are the marginal p.d. functions given by Eqs. ( 7 ) and (8).2.9. Regression lines
The regression of )( on
)I
is given bylCo
X/(x
J fI )
cI~
fOOl
(~j/!)
dOK
The expression which results after substitution of Eq.
(4)
has been evaluated using equations 11.4.28 and 13.1.27 from Abramowitz and Stegun ...(1965), with the result
(32 )
in which
11
is the confluent hypergeometric function, and Z is a coefficient given byIt appears that in general the regresswn of
X
onY
is non-linear.Exception should be made for the two trivial cases of stochastic independence
and linear dependence of )( and
)I.
In the case of stochastic independence,
~
=
0, substitution of whichgives as expected.
X=/
;)
J
I
In the case of linear dependence of
X
andy,
-I
= 1. As-i
...
1 ,~
-+
-
00. Using asymptotic expansions of the confluent hypergeometricfunction, it may be shown that
Therefore,
as expected.
x_
l
if (36)The regression of
Y
onX
is obtained by interchanging~
.
and/f-
inthe preceding expressions, because the bivariate Rayleigh distribution is
symmetric with respect to ~ and ;J-.
References
Abramm·Titz, M. and Stegtul, LA. : "Handbook of Mathematical Functions",
Dover Publications, Inc., NevT York, 1046 pages, 1965.
Bretschneider, C.L.: "Wave Variability and Wave Spectra for Wind-generated
Gravity Waves", Techn. Mem. 118, B.E.B., Dept. of the Army, 19)2 pages, 1959.
Bretschneider, C.L.: "Generation of Waves by Wind; State of the Art", NESCO
Report SN- 134-6, 96 pages, 1965.
Lighthill, M.J.: "Fourier Analysis and Generalised Functions", Cambridge
Univ. Press, 79 pages, 1966.
LongueJi-Higgins, M.S. (1952); "On the Statistical Distribution of the Heights
of Sea Waves", Journal of Marine Research, vol. 11, No.3, p. 245 - .266.
Longuet-Higgins, M.S.: "The distribution of intervals betw·een zeros of a
stationary random ftulction", Phil. Trans. Roy. Soc., vol. 254, No. 1047,
p. 557-599, 1962.
Magnus, W. and Oberhettinger, F.: "Formeln tuld Satze fUr die speziellen Funktionen
der liathematischen Physik", Berlin, 172 p., 1943.
Middleton, D.: "Statistical Communication Theory", Mc. Graw-Hill Book Company,
Inc. , New York, 1140 pages., 1960 , ~ .. -~ _ ... - ~.
Nakagami, M.: "On the Intensity Distribution .• and its Application to Signal
Statistics", Radio Science, vol. 68D, No.9, p.995-1003, sept. 1964.
Putz, R.A.: "Statistical Distribution for Ocean Waves;' Trans. A.G.U., Vol. 33,
No.5, p. 685-692, 1952.
Rice, S.O.: "Mathematical Analysis of Random Noise", The Bell System Techn. Jour.,
Vol. 23, p. 282-332, 1944; Vol. 24, p. 46-156, 1945.
Uhlenbeck, G.E.: "Theory of Random Processes", MIT Radiation Lab. Rept. 454, 1943.
Frechet, M.: "Sur les tableaux de correlation dont les marges sont donnees",
Annales de l'Universite de Lyon, Section A, p. 53-77, 1951.
Frechet, M.: "Sur les tableaux de correlation dont les marges sont donnees",