0% found this document useful (0 votes)
13 views8 pages

Truncated Cauchy Distribution

The document summarizes the truncated Cauchy distribution. It provides the probability density functions (pdfs) and cumulative distribution functions (cdfs) for the right-truncated, left-truncated, and doubly truncated Cauchy distributions. Key points include that the pdf of the doubly truncated distribution involves the difference of two inverse tangent terms with arguments depending on the truncation points, and its cdf involves taking the inverse tangent of the fraction of the argument minus the location parameter over the scale parameter.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
13 views8 pages

Truncated Cauchy Distribution

The document summarizes the truncated Cauchy distribution. It provides the probability density functions (pdfs) and cumulative distribution functions (cdfs) for the right-truncated, left-truncated, and doubly truncated Cauchy distributions. Key points include that the pdf of the doubly truncated distribution involves the difference of two inverse tangent terms with arguments depending on the truncation points, and its cdf involves taking the inverse tangent of the fraction of the argument minus the location parameter over the scale parameter.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 8

TRUNCATEDCAUCHYDI

STRI
BUTI
ON

I
ntr
oduct
ion

The Cauchy di str


ibuti
on,whi ch fi
rstappear ed in 1853,i sav ery signif
icantcontinuous
probabi l
i
tydistri
bution(Johnsonetal .
,vol-
1,2004).I
twast horoughl
yinvestigatedbynumer ous
researchers(Johnsonetal .,vol
-1,pp289–329, 2004).Ithasbeenusedi nmanydi ff
erentareas,
i
ncl udi
ngmechani calandel ect
ri
caltheory,physi
calanthropology,measurementandcal i
brati
on
i
ssues,model l
ingdept hmapdat a,pri
cingspeculati
veasset slikest
ockr eturns,andthephase
der i
vati
veofai rcomponent sinur banenv ir
onment s( NadarajahandKot z,2007a).Cauchy
distri
butionisjustaspopul arasthenor maldistr
ibut
ion, mostlyduetoit
smor eusefulnessasa
priordistri
buti
onandmor er eal
i
stictai
ls(Nadaraj
ahandKot z,2007b).

 Thepdfofr
ightt
runcat
eddi
str
ibut
ion:

f(x) f(
x)
f
(x)= B = -
∞ <X<B
∫A
f(
x )
dx F(
B)

Thecor
respondi
ngcdfi
s:
x
H(
X)= ∫h(x)dx= FF((Bx))
-

x<B

 Thepdfofl
eftt
runcat
eddi
str
ibut
ion:

f
(x) f
(x)
(
fx)= ∞ = A<X<∞
∫A
(
fx)
dx 1-
F(A)

Thecor
respondi
ngcdfi
s:
x
F(x)-
F(A)
H( ∫
X)= h(
x)
A
dx=
1-F(A)
A<X<B

 Thepdfofdoublet
runcat
eddi
str
ibut
ionoft
hel
eftt
runcat
ionpoi
ntAandr
ightt
runcat
ion
poi
ntBisgivenby:

f(x) f(
x)
f
(x)= B = A<X<B
∫A
f(
x d
)x F(
B)-F(
A)

Thecor
respondi
ngcdfi
s:
x
F(
x)-
F(A)
H( ∫
X)= h(
x)
A
dx=
F(
B)-
F(A)
A<X<B

ThepdfofCauchyDi
str
ibut
ioni
sgi
venby
:

1 1
(
fx)= -
∞ <X<∞ λ>0 -
∞ <θ<∞
λπ x-θ2
1+( )
λ

Wher
eθandλar
ethel
ocat
ionandscal
epar
amet
err
espect
ivel
y.
Doubl
eTr
uncat
edDi
str
ibut
ion:
Thepdfofdoubl
etr
uncat
eddi
str
ibut
ion.

1 1
λπ x- θ2
1+( )
λ
f
(x)= B A<X<B λ>0 -
∞ <θ<∞
1 1

Aλπ x- θ2
dx
1+( )
λ

Takedenomi
nat
or,

1 1

B
dx
λπ A θ2
x-
1+( )
λ
x-
θ
Let =u,x-
θ=λu,
dx=λdu
λ
A-
θ B- θ
<u<
λ λ
B-
θ

=∫λ1π1+1(u) λdu
λ
A-
θ
λ
2

1 B-
θ A-
θ
= λtan-1( ) -
tan-1( )
λπ λ λ

Nowputt
hisr
esul
tint
hepdfofDoubl
etr
uncat
eddi
str
ibut
ion:

1 1
λπ x- θ2
1+( )
λ
f
(x)= A<X<B
1
λπ
λ[
t
B-
θ
an-1( )
λ
-an-1
t ( )
A-
θ
λ
]

1 λ
f
( =
x) A<X<B
( )
2
1 B
- -
θ 1 A
- -
θ 1+(
x-θ)
an ( )
t -
tan
λ λ
Thi
sist
hepdfofDoubl
eTr
uncat
edDi
str
ibut
ion.

CDFOFTRUNCATEDDI
STRI
BUTI
ON
1 λ x-
θ
2 dx
x
CDF=∫ l
et =u,x-
θ=λu,
1 B-
θ 1 A-
( )
θ 1+(
x-θ) λ
A
- -
t
an ( ) -
tan
λ λ
dx=λdu
x-
θ
1 1
2du
=∫λ
1 B-
θ 1 A-
θ 1+(
( )
A-θ
- - u)
λ t
an ( ) -
tan
λ λ

F(
X)=
t
X-
θ
an-1( )
λ
-
tan-1 (Aλ-θ)
[t
B-
θ
an-1( )
λ
t
-an-1 (Aλ-θ)]
PROPERTI
ESOFDTCD

1:Mean

x λ
2dx
B
MeanE(
X)=∫
1 B-
θ 1 A-
( )
θ 1+(
x-θ)
A
- -
an ( )
t -
tan
λ λ
x-
θ
Let =u,x-
θ=λu,
dx=λdu x=ϑ+λu
λ
B-
θ
ϑ+λu 1
∫λ
2du
1 B-
θ 1 A-
θ 1+(
u)
( )
A-
θ
- -
an ( )
t
λ -
tan
λ λ

B-
θ
ϑ 1 B-
θ
λμ 1
2du 2du
=∫λ +∫λ
B-
θ
( )
A-
θ
( ) B-
θ A-
θ
A-θ A-θ
λt -
1 -
1 1+(
u) -
1 -
1 1+(
u)
an ( ) -
tan an ( )
λt -
tan
λ λ λ λ

an ( )
2 2
B-θ-
1 A-θ -
1 B-θ A-
θ
ϑ an ( )
t -t ) λl
n (
1+ ) -
λln (
1+
λ λ λ λ
= +
an ( ) 2t an ( )
-
1B-
θ A-θ -
1 B-θ A-θ-
1 -
1
t
an ( ) -t an ( ) -t
λ λ λ λ
2 2
B-
θ A-
θ
λln [(1+ )/ (1+ )]
λ λ
Mean=θ+
2t
B-
θ
an-1( )
λ
-
tan-1
A-
λ
θ
( )
2:Var
iance=E(
X2)
-[E(
X)2
]
B-
θ
( 2
ϑ+λu) 1
E(
X)=
2
∫ -1 B-θ -1 A-θ 1+(u)2 du
λ

( )
A-
θ
t
an ( )
λ -
tan
λ λ
=
B-
θ
( 2
ϑ) 1 B-θ
( 2
λu) 1 B-θ
2λθ
∫λ du+∫λ du+∫λ

( ) ( )
2 2
A-θ
B-
θ A-
θ 1+(
u) A-θ
B-
θ A-
θ 1+(
u) A-θ
B-
θ
an-1( )
λt -an-1
t λt an-1( ) -an-1
t λt an-1( ) -an-1
t
λ λ λ λ λ
=

( 2
ϑ) B-
θ
an ( )
t -
tan ( )+
A-
θ
-
1
( λ λ)
B-θA+θ
-
-
1
λ2
an ( ) an ( )
B-
θ-
1 A-
θ -
1 λ λ B-
θ A-
θ -
1 -
1
t
an ( ) -
t an ( )
t -
t
λ λ λ λ
λ2+(B- θ)2
λθln (2
( )
2
1 B
- -
θ 1 A
- -
θ λ+( A- θ)
[
-an ( )+t
t an ]
+
λ λ
t
B-
θ
an-1( )
λ
-tan-1
A-
λ
θ
( )
λ2+(B-
θ)2
λ(
B-A) +λθln (2 2
λ+( A-
θ)
E(
X)
2
=( 2
ϑ)- ( 2
λ)+
t
B-
an-1( )
λ
θ
-an-1
t
A-
λ
θ
( )
Var
iance=E(
X2)
-[E(
X)2
]

( ]
2
λ2+( 2 2 2
B-θ) B-
θ A-
θ
λ(
B-A)+λθln (2 2 λl
n [
(1+ )/1+( )]
λ+(A-θ) λ λ
Var
iance=( 2
ϑ)- ( 2
λ)+ -
(θ+ )
t 1 B
- -
an ( )
λ
θ
-t
an-1 A
λ
-
θ
( )
2t 1 B
- -
θ
an ( )
λ
t
-an1 A
- -
θ
λ ( )
( ]
2 2 2
B-
θ A-
θ
λl
n [
(1+ )/1+( )]
λ(
B-A) λ λ
Var
iance= -
( 2
λ)[(
1+ )]
t 1 B
- -
θ
an ( )
λ
-
tan1 A
-
λ
-
θ
( ) 2t 1 B
- -
θ
an ( )
λ
-
tan1 A
- -
θ
λ ( )
Andst
andar
ddev
iat
ioni
nequal
tot
hesquar
erootofv
ari
ance.

4:Recur
rencer
elat
ionf
orMoment
s:
Ther
thmomentaboutt
hepoi
ntX=θi edbyu'
sdenot (
r
θ)andi
sgi
venby


B
r
u'
(
r
θ)= (
X-θ)f
(x)
dx
A


B
r
u'
(
r
θ)= (
X-θ)f
(x)
dx
A

1 1(
X-
r
θ)

B
dx
A
t 1 B
- -
θ
an ( )
λ
-
tan1 A
-
λ
-
θ
( ) ()
λ[
θ2
1+x- ]
λ

1 B-
θ
u)
r λ (
r
= λ)∫
( 2du
1 B
- -
θ
an ( )
t
λ
-
tan1 A
-
λ( )
-
θ Aλ-θ1+(u)

B-
θ
1 r λ( 1+u2)
(u)r
-2
-(
u)r
-2
= ( ∫
λ) A-θ du
( )
2
1 B
- -
θ 1 A
- -
θ 1+(u)
tan ( ) -
tan λ
λ λ

B-A
u'
(θ)
=λ -
( 2 '
λ)ur-2(
θ)
1 B-
θ 1 A-
θ
( )
r
- -
(
r-
1)an ( )
t -
tan
λ λ
Putr
=2wehav
e:

B-
A
u'(θ)
=λ -
( 2
λ)
B-
θ
( )
A-
θ
2

an-1( )
t t
-an-1
λ λ
5:Mode:
TheDTCDi
suni
modal
andt
hemodeoccur
satt
hepoi
ntX=θ.
'
Modei
sthatv
alueofxf
orwhi
chf
(x)i
smaxi
mum t
hati
smodei
sthesol
uti
onoff
(x)=0,
'
'
pr
ovi
dedt
hatf(
x)<0.Fort
hepdfofDTCDt
hemodei
s:
'
(
fx)=0

x=θ

(
'
'
Andf
x)
=-
2/
{
[t
an ( -
1 B-
θ
λ
)
-an-1
t ( )]
A-
θ 3
λ
}
(λ) <0

6:Quar
ti
le:
I
fQpi
sthept
hquar
ti
le,
thenF(
Qp)=pi
mpl
i
es,

1 1
dx=p
( )[ ( )]
Qp
F( x≤Qp)=∫
Qp)=p( 1 B
- -
θ 1 A
- -
θ θ2
an ( )
t -
tan λ1+x-
A
λ λ λ

Qp-
θ
1 1
= ∫ -1 B-θ -1 A-θ 1+(u)2 du=p
λ

( )
A-
θ
t
an ( )
λ t
-an
λ λ

=
t
Q-
λ
θ
an-1( p )-tan-1 ( )
A-
θ
λ
=p
t -1 B
an ( )
λ
-
θ
-t
an 1 A
-
( )
λ
-
θ

Qp =θ+λt
an [
p(t
an ( -
1 B-
θ
λ
)
-t ( )
an-1
A-
θ
λ
+t ( )
an-1
A-
θ
λ
)
]

I
fp=0.
5,t
henwegett
hemedi
an.

Q0.5=θ+λt
an [
0. t
5(an ( -
1 B-
θ
λ
)
-tan-1( )
A-
θ
λ
+t ( )
an-1
A-
θ
λ
)
]

Thi
sist
hemedi
anoft
het
runcat
edCauchydi
str
ibut
ion.

I
fweputp=0.
25and0.
75t
henwegett
hef
ir
standt
hir
dquar
ti
le.
Char
act
eri
sti
csFunct
ion:

7:Sur
viv
alFunct
ion:1-
F(x)

Sur
viv
alf
unct
ion

8:Hazar
dFunct
ion:f
(x)
/sur
viv
alf
unct
ion

H.
F:=

9:Cumul
ati
veHazar
dfunct
ion:
Par
amet
erEst
imat
ion
Maxi
mum Li
kel
ihoodMet
hod
Foranygi
venθandλt
hev
alueofLwi
l
lbemaxi
mum wi
thr oAandB
espectt

Whent
B-
θ
an-1( )
λ
t
-an-1
A-
θ
λ( )
i
smi
nimum.Denot
ingt
heor
derst
ati
sti
cs

Cor
respondi
ngtothesampl
ebyA<x1<x2<…<xn<B,theaboveexpr
essi
onwill
bemi ni
mum
̂ ̂
whenA=x 1
andA=xn.Hencet
heMLEsofAandBareA =x1andB =xnrespect
ivel
y.Now,
the
l
og-l
i
keli
hoodfunct
ioni
s:
n

l
ogL=-
nlogt
B-
θ
an ( )
λ[-
t -
1
an-1
A-
θ
λ
+nl( )]
ogλ- l λ2+(
og { x- 2
θ)}∑ i
-1

Thecor
respondi
ngmaxi
mum l
i
kel
i
hoodequat
ionsar
e,

dl
ogL
dx
=-
nlogt -
[
1 B-
θ
an ( )
λ
-
tan-
( )]
1 A
λ
θ -1
-
[2
-
λ
λ+(
B- 2+ 2
θ) λ+(
λ
A- 2]
θ)
+∑n
i
-
2(
1 2
λ+(
x-θ)
x-
θ)2 =0

[ an ( )
λ ] λ+(
-
1
dl
ogL B-
θ -
1 A-
θ B-
-
1θ A-θ n 2λ n
=-
nlogt
an ( ) -
t [ + 2 ]
+ -
2∑ 2 = 2 1 2 2
dx λ B-θ) λ+(A-θ) λ λ+(x
-θ) i
-

0
Theparameter
sθandλandmaybeest i
matednumerical
l
ybysol
vi
ngtheabovetwoequati
ons
usi
ngit
erat
iveprocedures,
suchastheNewtonRaphsonmet
hod,wi
thsuf
fi
cient
lycl
oseini
ti
al
val
uessothattheprocedureconv
erger
api
dlytosol
uti
on.

Ref
erences

1. Nadar
ajah,
S.,
&Kot
z,S.(
2007)
.Skewmodel
sI.
 Act
aappl
i
candaemat
hemat
icae,
98(
  1),
1-28.

2. Nadar
ajah,
S.,
&Kot
z,S.(
2006)
.At
runcat
edCauchydi
str
ibut
ion.

3. Ar
nol
d,B.C.
,&Beav
er,
R.J.(
2000)
.Theskew-
Cauchydi
str
ibut
ion.
 St
ati
sti
cs&pr
obabi
l
ity
l
ett
ers,
49(
  3),
285-
290.

You might also like