WEIBULL DISTRIBUTION

CHARUDARSAN P
Department Of Statistical Sciences
Kannur University
Kerala 670567
India
email:charudarsanp@gmail.com

Introduction
The weibull distribution is named after the Swedish physicist, Walodi Weibull,
who (1930) used it to represent the distribution of the breaking strength of
material and in 1951 for a wide variety of other applications. Together with,
the normal, exponential, chi-square, t and F distributions the weibull distri-
bution is without any doubt the most popular model in modern statistics. It
is of almost interest to theory-oriented statisticians because of its great num-
ber of special features and to practitioners because of its ability to fit to data
from various fields, ranging from life data to weather data or observations
made in economics and business administration, in hydrology, in biology or
in the engineering science.

History
The history and discovery of what is now known as the weibull distribution
is even more exciting than that of the normal distribution, which lasted 60
to 70 years. The key events in the derivation of the weibull distribution
took place between 1922 and 1943. This process is so astonishing because
there were three groups of persons working independently with very different
aims, thus forming a chain of three links. WALODDI WEIBULL (1887-
1979) is the last link of this chain. The distribution bears his name for good
reasons because it was he who propagated this distribution internationally
and interdisciplinarily. Two of the approaches leading to this distribution
are located in engineering practice and are more or less heuristic whereas the
third approach, the oldest and purely scientific one, is based on theoretical
reasoning and is located in research on statistical methods and probability
theory.

1
Probability density function
If the random variable
c
X −µ

Y =
λ
Has the exponential distribution with p.d.f

g (y) = e−y , y > 0 (1)

Then random variable X has Weibull distribution with parameters c > 0, λ

> 0 and µ
The p.d.f of X is
c−1 " c #
c X −µ X −µ
 
f (X) = exp − ,x > µ (2)
λ λ λ

Cumulative distribution function

The probability mass function of weibull distribution is
c−1 " c #
Z x
c X −µ X −µ
 
F (X) = exp − dx// (3)
0 λ λ λ

By substituting c
x−µ

u=
λ
we get
c 
x−µ
 
F (x) = 1 − exp − (4)
λ
This form of cumulative distribution function makes the weibull distribution
more suitable to be used as a tolerance distribution

Standard weibull distribution function

Putting λ = 1 and µ = 0 we have
c
g (x) = cxc−1 e−x , x > 0 (5)

This density function is the standard Weibull distribution with single param-
eter c

2
Survival function
The survival function of weibull distribution is
c 
x−µ
 
S (x) = 1 − F (x) = exp − − (6)
λ

Hazard rate function

The hazard rate function of weibull distribution is
c−1
f (x) c x−µ
 
h (x) = = (7)
S (x) λ λ

Moments
General expression for row moments of X
c−1 " c #
Z ∞
c X −µ X −µ
  
F (X) = Xr exp − dx (8)
µ λ λ λ
X−µ c
 
By substituting u = λ
Z ∞ r
1
E (X r ) = λµ c = µ e−µ dµ (9)
0

r
1
 
(rn ) µr−n λn Γ
X
= +1 (10)
n=0 c

for standard weibull distribution we have

Y = X c [λ = 1, µ = 0]
1
X=Y c

The r-th moment about origin is

Z ∞
r
 
′ r
−y
µr = y e dy = Γ
c +1 (11)
0 c
In particular
1
 
′
µ1 = Γ +1 (12)
c

3
 2
 
′
µ2 = Γ +1 (13)
c

3
 
′
µ3 = Γ +1 (14)
c

4
 
′
µ4 = Γ +1 (15)
c

2
2 1
2    
′ ′

µ2 = µ1 − µ2 =Γ +1 − Γ +1 (16)
c c
for large c

π2
!
γ 1 1
E (X) = 1 − + + γ 2 2 = 1 − 0.577c−1 + 0.98905c−2 (17)
c 2 6 c

π2 1.6449
V (X) = 2
= (18)
6c c2
Here γ is Euler’s constant

4
Shape parameter c
The parameter c is called the shape parameter of weibull distribution. The
shape of the distribution changes rapidly with the change in c.In Figure 1
both λ and µ are kept constant and c is increased

Figure 1: Shape parameter

5
Scale parameter λ
If both c and µ are kept constant and λ is increased the distribution gets
stretched out to right

Figure 2: Scale parameter

6
Properties
For c > 1 the p.d.f tends to zero as x tends to 0 and there is a single mode
at
" #1
(c − 1) c
x=λ =µ (19)
c

this value tents to(µ + λ) very rapidly as c tents to ∞.

For 0<x<1 the mode is at zero, and f(x) is a decreasing function of x for all
x> µ. the median is
1
λ [log2] c + µ √
For c=3.6 (approx) β is zero and β2 = 2.72. Thus for values of c in the
neighbourhood of 3.6 the distribution is similar in shape to a normal distri-
bution. For c>3.6 distribution becomes negatively skewed and platikurtic.
For c<3.6 the distribution is positively skewed.

Relation between weibull and other distribu-

tions
* Weibull exponential
X ∼ W (c, λ) iff X c ∼ exp (λc )
Weibull distribution in (1) when c = 1 becomes two parameter exponential
distribution

x−µ
F (x) = 1 − e−( λ ), x > µ (20)

When c = 2 weibull distribution in (2) becomes two parameter rayleigh

distribution
x−µ 2
F (x) = 1 − e−( λ ) ,x > µ (21)
h i
x−µ
If X has a weibull distribution in (1) then the pdf of Y = −log λ
is
given by

PY (y) = exp (−ey ) e−y ; −∞ < y < ∞ (22)

Which is one of the extreme value distribution

7
Applications
Because of its ability to fit to data from various fields the weibull distribution
has many applications
⋆ Applications in reliability
Yield strength of steel, fatigue life of steel,Fracture strength of glass,Pitting
corrosion in pipes,Adhesive wear in metals,Failure of carbon-fiber compos-
ites,Failure of coatings...
⋆Applications on geophysics
Wind-speed data analysis,Earthquake magnitude,Volcanic occurrence data,Low-
flow analysis
⋆Food science
Sterility in thermal preservation method
⋆Social science
Unemployment duration
⋆Environment
radioactivity
⋆Medical science
Survival data
⋆ect ...

Bibliography
⋆KC Bhuyan”Probability Distribution Theory and Statistical Inference”.
⋆Horst Rinne”The Weibull Distribution: A Handbook ”.
⋆Chin-Diew Lai”Weibull Distributions and Their Applications”.
⋆Jaishna M V ”A study on some generalization of weibull distribution(project
report Kannur university) ”.
⋆ Sudha G Purohit, Shylaja R Deshmukh and Sharad D Gore ”Statistics
Using R”.