Scribd
Upload
47 views

Derivation of The T-Distribution: T U V/N

The t-distribution was introduced in 1908 by William Sealy Gosset and is defined using variables from the standard normal distribution and the chi-squared distribution. The probability density function of the t-distribution can be written in terms of the gamma function. For large degrees of freedom, the t-distribution approaches the standard normal distribution, but it has fatter tails for low degrees of freedom, making it useful for statistical inference involving small sample sizes.

Uploaded by

AYU JANNATUL
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
47 views

Derivation of The T-Distribution: T U V/N

The t-distribution was introduced in 1908 by William Sealy Gosset and is defined using variables from the standard normal distribution and the chi-squared distribution. The probability density function of the t-distribution can be written in terms of the gamma function. For large degrees of freedom, the t-distribution approaches the standard normal distribution, but it has fatter tails for low degrees of freedom, making it useful for statistical inference involving small sample sizes.

Uploaded by

AYU JANNATUL
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Derivation of the t-Distribution

Shoichi Midorikawa

Student’s t-distribution was introduced in 1908 by William Sealy Goset. The


statistc variable t is defined by
u
t= √ ,
v/n

where u is a variable of the standard normal distribution g(u), and v be a


variable of the χ2 distribution Tn (v) of of the n degrees of freedom. Thus, we
can express the distribution function of t in terms of g(u) and Tn (v).
The distribution function g(u) and Tn (v) are represented by
1
g(u) = √ e−u /2
2
(1)
2π
and
1
Tn (v) = v (n−2)/2 e−v/2 (2)
2n/2 Γ(n/2)
respectivley. √
The probabilistic function fn (t) of t = u/ v/n can then be written as
∫ ( )
u
fn (t) = δ t − √ g(u)Tn (v) dudv,
v/n
( √ )
where δ t − u/ v/n is a Dirac’s delta function.
√
Before integrate over u, we introduce a variable y = u/ v/n, and integrate
over y instead of u.
∫ √ √
v
fn (t) = δ (t − y) g( v/n y)Tn (v) dydv
n
∫ √
v √
= g( v/n t)Tn (v) dv
n
Now, substituting eq.(1) and eq.(2) into this equation, we get
∫ ∞
1
v (n−1)/2 e−(1+t /n)v/2 dv
2
fn (t) = √
n/2
2πn 2 Γ(n/2) 0
( )
t2 v
Furthermore, we rewrite the equation using a variable x = 1 + in-
n 2
stead of v, to find that
∫
(1 + t2 /n)−(n+1)/2 ∞ (n+1)/2−1 −x
fn (t) = √ x e dx.
πn Γ(n/2) 0

1
The integrand of the right hand side can be represented by using the gamma
function, ∫ ∞
Γ(α) = xα−1 e−x dx.
0
Thus we obtain
( )−(n+1)/2
Γ(n + 1)/2) t2
fn (t) = √ 1+
πn Γ(n/2) n

Furthermore, gamma functions are repsented by beta funtion as


√
Γ(1/2) Γ(n/2) π Γ(n/2)
B(1/2, n/2) = = .
Γ((n + 1)/2) Γ((n + 1)/2)

So we finaly otain the studen’s t-distribution function as follows,


( )−(n+1)/2
1 t2
fn (t) = √ 1+
π B(1/2, n/2) n
.
We show the t distribution function as Figure 1. For comparison, we also
show the standard normal distribution function (N (0, 1) - solid line).

0.4 N(0, 1)
n= 1
n= 3
n=10
n=30

0.3

0.2

0.1

0
-4 -2 0 2 4

Figure 1: tdistribution function y = fn (t) of n = 1, 3, 10, 30 degrees of freedom,


and the standard normal distribution functionN (0, 1)as solid line.

You might also like