Log-normal distributions

Vijay Kumar

Department of Mathematics and Statistics

DDU Gorakhpur University, Gorakhpur
[email protected]

Feb. 2023
 Log-noraml PDF Median Mode Moments Examples

Log-normal distribution:

If Y = log X ∼ N (µ, σ 2 ) then X ∼ LN µ, σ 2 .

(   )
1 1 y−µ 2
dF (y) = √ exp exp − dy
σ 2π 2 σ

y = log x, dy = x1 dx, y = −∞, x = 0, y = ∞, x = ∞

(   )
1 1 log x − µ 2
dF (x) = √ exp − dx
σx 2π 2 σ

Therefore, (   )
1 1 log x − µ 2
f (x) = √ exp − ; x>0
σx 2π 2 σ

V. Kumar, DDU Gorakhpur University, Gorakhpur B.Sc.-I : Log-normal distribution February, 2023 2/7
 Log-noraml PDF Median Mode Moments Examples

Median:

M 1
∫ f (x) dx =
0 2
(   )
1 M1 1 log x − µ 2 1
√ ∫ exp exp − dx =
σ 2π 0 x 2 σ 2

log x − µ 1
= z, dx = dz,
σ xσ
log M − µ
x = 0, z = −∞, x = M, z =
σ
log M −µ
1 σ 2 1
√ ∫ e−z /2 dz =
2π −∞ 2
0 1 2 logM −µ
We know that √1 ∫
2π −∞
e− 2 z dz = 12 . Therefore, σ = 0 ⇒ M = eµ . Median at the point
eµ .
V. Kumar, DDU Gorakhpur University, Gorakhpur B.Sc.-I : Log-normal distribution February, 2023 3/7
 Log-noraml PDF Median Mode Moments Examples

Mode:
f ′ (x) = 0
" (   )#
d 1 1 log x − µ 2
√ exp − =0
dx σx 2π 2 σ
(   )
1 1 log x − µ 2
− √ exp −
σx2 2π 2 σ
(   )
1 log x − µ 2
  
1 2 log x − µ 1
+ √ exp − − =0
σx 2π 2 σ 2 σ xσ
(   )
1 log x − µ 2

1 log x − µ
=− √ exp − 1+ =0
σx2 2π 2 σ σ2
f (x) log x − µ + σ 2
 
=0
x σ2
2
⇒ log x − µ + σ 2 = 0 or x = eµ−σ

V. Kumar, DDU Gorakhpur University, Gorakhpur B.Sc.-I : Log-normal distribution February, 2023 4/7
 Log-noraml PDF Median Mode Moments Examples

Moments:

Z∞ (   )
1 1 r 1 log x − µ 2
µ′r r r
= E (X ) = ∫ x f (x) dx = √ .x exp −
σ 2π x 2 σ
0

log x − µ
=z ; x = eµ+σz
σ
dx = σeµ+σz dz ; x = 0, z = −∞, x = ∞, z = +∞.

1 −∞ (µ+σz)r −z 2 /2
µ′r = √ ∫ e e dz
2π 0
eµr −∞
 
1 2 2 2

1 2 2
= √ ∫ exp − z − 2σrz + σ r + σ r .dz
2π 0 2 2
 
1 2 2 1 −∞ 1
= eµr+ 2 σ r √ ∫ exp − (z − σr)2 .dz
2π 0 2

V. Kumar, DDU Gorakhpur University, Gorakhpur B.Sc.-I : Log-normal distribution February, 2023 5/7
 Log-noraml PDF Median Mode Moments Examples

Moments cond...
 
1 2 2
µ′r == exp µr + σ r
2
 
1
µ′1 = exp µ + σ 2
2
′
µ2 = exp 2µ + 2σ 2

2 2 2
µ2 = µ′2 − µ′1 = e2µ+2σ − e2µ+σ
2
 2 
µ2 = e2µ+σ eσ − 1
2
Mode = eµ−σ
Median = eµ
Mean > Median > Mode
Positively skewed and leptokurtic

V. Kumar, DDU Gorakhpur University, Gorakhpur B.Sc.-I : Log-normal distribution February, 2023 6/7
 Log-noraml PDF Median Mode Moments Examples

Examples:

If X1 ∼ LN µ1 , σ12 and X2 ∼ LN µ2 , σ22 , X1 and X2 are independent then

X1 X2 ∼ LN µ1 + µ2 , σ12 + σ22

X1 2 2


X2 ∼ LN µ1 − µ2 , σ1 + σ2
Since X1 ∼ LN µ1 , σ12 and X2 ∼ LN µ2 , σ22 , we have

X1 ∼ LN µ1 , σ12 ; Y1 = log X1 ∼ N µ1 , σ12

X2 ∼ LN µ2 , σ22 ; Y2 = log X2 ∼ N µ2 , σ22

Y1 + Y2 = log X1 + log X2 ∼ N µ1 + µ2 , σ12 + σ22

log (X1 X2 ) ∼ N µ1 + µ2 , σ12 + σ22

log (X1 X2 ) ∼ LN µ1 + µ2 , σ12 + σ22

V. Kumar, DDU Gorakhpur University, Gorakhpur B.Sc.-I : Log-normal distribution February, 2023 7/7