From Wikipedia, the free encyclopedia
Finished writing a draft article? Are you ready to request review of it by an experienced editor for possible inclusion in Wikipedia? Submit your draft for review!
The studentized range is defined as
q
=
n
r
2
S
{\displaystyle q={\frac {{\sqrt {n}}r}{{\sqrt {2}}S}}}
. To create the studentized range PDF, we will need to integrate over all combinations of a range distribution, and the
χ
{\displaystyle \chi }
distribution which would give a value of q .
The range distribution is:
f
(
r
;
k
)
=
k
(
k
−
1
)
∫
−
∞
∞
f
(
u
−
r
)
f
(
u
)
[
F
(
u
)
−
F
(
u
−
r
)
]
k
−
2
d
u
{\displaystyle f(r;k)=k(k-1)\int _{-\infty }^{\infty }f(u-r)f(u)\left[F(u)-F(u-r)\right]^{k-2}\,{\text{d}}u}
The f and F are the averages of n standard normal variables, which is
f
(
z
)
=
n
ϕ
(
n
z
)
{\displaystyle f(z)={\sqrt {n}}\phi ({\sqrt {n}}z)}
, and
F
(
z
)
=
ϕ
(
n
z
)
{\displaystyle F(z)=\phi ({\sqrt {n}}z)}
. This gives us:
f
(
r
;
k
)
=
k
(
k
−
1
)
∫
−
∞
∞
n
ϕ
(
n
(
u
−
r
)
)
n
ϕ
(
n
u
)
[
Φ
(
n
u
)
−
Φ
(
n
(
u
−
r
)
)
]
k
−
2
d
u
{\displaystyle f(r;k)=k(k-1)\int _{-\infty }^{\infty }{\sqrt {n}}\phi ({\sqrt {n}}(u-r)){\sqrt {n}}\phi ({\sqrt {n}}u)\left[\Phi ({\sqrt {n}}u)-\Phi ({\sqrt {n}}(u-r))\right]^{k-2}\,{\text{d}}u}
if
t
=
n
u
{\displaystyle t={\sqrt {n}}u}
then
d
t
=
n
d
u
{\displaystyle {\text{d}}t={\sqrt {n}}{\text{d}}u}
f
(
r
;
k
)
=
k
(
k
−
1
)
n
∫
−
∞
∞
ϕ
(
t
−
n
r
)
ϕ
(
t
)
[
Φ
(
t
)
−
Φ
(
t
−
n
r
)
]
k
−
2
d
t
{\displaystyle f(r;k)=k(k-1){\sqrt {n}}\int _{-\infty }^{\infty }\phi (t-{\sqrt {n}}r)\phi (t)\left[\Phi (t)-\Phi (t-{\sqrt {n}}r)\right]^{k-2}{\text{d}}t}
Combining this with the
χ
{\displaystyle \chi }
distribution gives:
f
(
q
;
k
,
ν
)
=
1
2
ν
2
Γ
(
ν
2
)
∫
0
∞
x
ν
2
−
1
e
−
x
2
k
(
k
−
1
)
n
∫
−
∞
∞
ϕ
(
t
−
n
r
)
ϕ
(
t
)
[
Φ
(
t
)
−
Φ
(
t
−
n
r
)
]
k
−
2
d
t
d
x
{\displaystyle f(q;k,\nu )={\frac {1}{2^{\frac {\nu }{2}}\Gamma \left({\frac {\nu }{2}}\right)}}\int _{0}^{\infty }x^{{\frac {\nu }{2}}-1}e^{-{\frac {x}{2}}}k(k-1){\sqrt {n}}\int _{-\infty }^{\infty }\phi (t-{\sqrt {n}}r)\phi (t)\left[\Phi (t)-\Phi (t-{\sqrt {n}}r)\right]^{k-2}{\text{d}}t{\text{d}}x}
or
f
(
q
;
k
,
ν
)
=
1
2
ν
Γ
(
ν
2
)
∫
0
∞
x
ν
−
2
e
−
x
2
k
(
k
−
1
)
n
∫
−
∞
∞
ϕ
(
t
−
n
r
)
ϕ
(
t
)
[
Φ
(
t
)
−
Φ
(
t
−
n
r
)
]
k
−
2
d
t
d
x
{\displaystyle f(q;k,\nu )={\frac {1}{{\sqrt {2}}^{\nu }\Gamma \left({\frac {\nu }{2}}\right)}}\int _{0}^{\infty }{\sqrt {x}}^{\nu -2}e^{-{\frac {x}{2}}}k(k-1){\sqrt {n}}\int _{-\infty }^{\infty }\phi (t-{\sqrt {n}}r)\phi (t)\left[\Phi (t)-\Phi (t-{\sqrt {n}}r)\right]^{k-2}{\text{d}}t{\text{d}}x}
Since x in the
χ
2
{\displaystyle \chi ^{2}}
distribution is defined as
x
=
ν
S
2
{\displaystyle x=\nu S^{2}}
we can set
y
=
2
ν
x
{\displaystyle y={\frac {\sqrt {2}}{\sqrt {\nu }}}{\sqrt {x}}}
to make
n
r
=
y
q
{\displaystyle {\sqrt {n}}r=yq}
and
d
y
=
d
x
2
ν
x
{\displaystyle {\text{d}}y={\frac {{\text{d}}x}{\sqrt {2\nu x}}}}
.
f
(
q
;
k
,
ν
)
=
ν
2
ν
−
1
Γ
(
ν
2
)
∫
0
∞
x
ν
−
1
e
−
x
2
k
(
k
−
1
)
n
∫
−
∞
∞
ϕ
(
t
−
n
r
)
ϕ
(
t
)
[
Φ
(
t
)
−
Φ
(
t
−
n
r
)
]
k
−
2
d
t
d
x
ν
2
x
{\displaystyle f(q;k,\nu )={\frac {\sqrt {\nu }}{{\sqrt {2}}^{\nu -1}\Gamma \left({\frac {\nu }{2}}\right)}}\int _{0}^{\infty }{\sqrt {x}}^{\nu -1}e^{-{\frac {x}{2}}}k(k-1){\sqrt {n}}\int _{-\infty }^{\infty }\phi (t-{\sqrt {n}}r)\phi (t)\left[\Phi (t)-\Phi (t-{\sqrt {n}}r)\right]^{k-2}{\text{d}}t{\frac {{\text{d}}x}{\sqrt {\nu 2x}}}}
f
(
q
;
k
,
ν
)
=
ν
ν
2
ν
−
1
Γ
(
ν
2
)
∫
0
∞
y
ν
−
1
e
−
ν
y
2
4
k
(
k
−
1
)
n
∫
−
∞
∞
ϕ
(
t
−
y
q
)
ϕ
(
t
)
[
Φ
(
t
)
−
Φ
(
t
−
y
q
)
]
k
−
2
d
t
d
y
{\displaystyle f(q;k,\nu )={\frac {{\sqrt {\nu }}^{\nu }}{2^{\nu -1}\Gamma \left({\frac {\nu }{2}}\right)}}\int _{0}^{\infty }y^{\nu -1}e^{-{\frac {\nu y^{2}}{4}}}k(k-1){\sqrt {n}}\int _{-\infty }^{\infty }\phi (t-yq)\phi (t)\left[\Phi (t)-\Phi (t-yq)\right]^{k-2}{\text{d}}t{\text{d}}y}