RANK CORRELATION COEFFICIENT
Syed . M Parveen
Dept.of Statistics
Maris Stella College
Vijayawada,AP.
Rank Correlation
Rank correlation was introduced by Spearman.
Let x1,x2,x3,…xn and y1,y2,y3,….yn be the ranks of n individuals of
the characteristics A and B respectively. The correlation obtained
for the ranks of individuals of two characteristics A and B is called
Rank Correlation.
Ranks of the individuals may or may not be the same.
Derivation of Spearman Rank Correlation Coefficient
Possibility 1. Ranks of the individuals are different ( not same).
Let the ranks of the individuals due to the characteristics A and B be 1,2,3,….n
(which are need not be in the same order) and in general xi ≠ y i .
x̅ = 1/n Ʃ xi ;
= 1/n *(1+2+3+….+n)
=n(n+1)/2n
=(n+1)/2 ………………………………..(1)
Similarly y̅ = (n+1)/2
x̅ = y̅
σx2 = var (x) = 1/n Ʃ xi
2 –( x̅)2
= 1/n * ( 12+22+32+……n2)-((n+1)/2)2
=(n(n-1)(2n-1)/6n)-((n+1)/2)2
=(n2-1)/12
Similarly σy2 = var (y) = (n2-1)/12
Let ρ be the rank correlation coefficient
ρ = Cov (X,Y)/ σ x *σ y
Cov (X,Y) = ρ * σ x * σ y
Let us consider deviation of the ranks as di
d i = xi – y i
d i = (xi- x̅)-( y i- y̅)
Squaring , summing over i from 1 to n and dividing with n, we
get
1/n Ʃ di
2 = 1/n Ʃ [ ( xi- x̅) – (y i- y̅)]2
=1/n Ʃ(xi- x̅)2 + 1/n Ʃ (yi- y̅)2 -2 *1/n Ʃ(xi- x̅)(yi- y̅)
=σx2+σy 2 -2*cov(X,Y) ……………………….[from equations (2), (4)]
=2* σx2 – 2 ρ * σx *σy (since σx2= σy 2 , σx = σy)
1/n* Ʃ di
2= 2* σx2( 1-ρ)
1 – ρ = (Ʃdi2) /2n σx2
= (Ʃdi2) / 2n((n2-1)/12)
= 6 (Ʃdi2) / n (n2-1)
ρ = 1 - 6 (Ʃdi2) / n (n2-1)

Derivation of rank correlation 1

  • 1.
    RANK CORRELATION COEFFICIENT Syed. M Parveen Dept.of Statistics Maris Stella College Vijayawada,AP.
  • 2.
    Rank Correlation Rank correlationwas introduced by Spearman. Let x1,x2,x3,…xn and y1,y2,y3,….yn be the ranks of n individuals of the characteristics A and B respectively. The correlation obtained for the ranks of individuals of two characteristics A and B is called Rank Correlation. Ranks of the individuals may or may not be the same.
  • 3.
    Derivation of SpearmanRank Correlation Coefficient Possibility 1. Ranks of the individuals are different ( not same). Let the ranks of the individuals due to the characteristics A and B be 1,2,3,….n (which are need not be in the same order) and in general xi ≠ y i .
  • 4.
    x̅ = 1/nƩ xi ; = 1/n *(1+2+3+….+n) =n(n+1)/2n =(n+1)/2 ………………………………..(1) Similarly y̅ = (n+1)/2 x̅ = y̅ σx2 = var (x) = 1/n Ʃ xi 2 –( x̅)2 = 1/n * ( 12+22+32+……n2)-((n+1)/2)2 =(n(n-1)(2n-1)/6n)-((n+1)/2)2 =(n2-1)/12 Similarly σy2 = var (y) = (n2-1)/12
  • 5.
    Let ρ bethe rank correlation coefficient ρ = Cov (X,Y)/ σ x *σ y Cov (X,Y) = ρ * σ x * σ y Let us consider deviation of the ranks as di d i = xi – y i d i = (xi- x̅)-( y i- y̅) Squaring , summing over i from 1 to n and dividing with n, we get 1/n Ʃ di 2 = 1/n Ʃ [ ( xi- x̅) – (y i- y̅)]2
  • 6.
    =1/n Ʃ(xi- x̅)2+ 1/n Ʃ (yi- y̅)2 -2 *1/n Ʃ(xi- x̅)(yi- y̅) =σx2+σy 2 -2*cov(X,Y) ……………………….[from equations (2), (4)] =2* σx2 – 2 ρ * σx *σy (since σx2= σy 2 , σx = σy) 1/n* Ʃ di 2= 2* σx2( 1-ρ) 1 – ρ = (Ʃdi2) /2n σx2 = (Ʃdi2) / 2n((n2-1)/12) = 6 (Ʃdi2) / n (n2-1) ρ = 1 - 6 (Ʃdi2) / n (n2-1)

Editor's Notes

  • #2 Syed . M Parveen Dept.of Statistics Maris Stella College(Autonomous) Vijayawada,AP.