Page 140 - DECO504_STATISTICAL_METHODS_IN_ECONOMICS_ENGLISH
P. 140

Statistical Methods in Economics


                   Notes          In the above illustration the numbers were chosen so that for an increase of 1 unit in the X series there
                                  is an increase of 2 units in the Y series. Thus the correlation is perfect and r equals +1. If the Y series
                                  had been 14, 12, 10, 8, 6 (the X series remaining the same) the value of r would have been —1. Thus —
                                  1 stands for perfect negative correlation, an increase in one series corresponding to a decrease in the
                                  other. It should also be noted in this connection that the coefficient of correlation (r) cannot be less
                                  than —1 nor more than + l.
                                  The above illustration suggests the question, “Will a linear relationship between X and Y always give
                                  perfect correlation ?”
                                  Assume the linear relationship
                                                     Y= aX + b

                                  Since               y = Y – M  and x =  X – M 2
                                                              2
                                                M + y = a(  +Mx  1 ) + b or y = ax
                                                  2
                                       (since –M – Mba  1  2 = 0)

                                                                                     ∑
                                                            ∑xy        ∑ax 2        ax 2
                                  and                 r =          =            =           =  ±  1
                                                                            2 2
                                                          ∑  2. ∑x  y 2  ∑  2.  ∑x  a x  2 ( a  2 ) ∑x  2
                                                       (The sign of r depends upon the sign of a.)
                                  Therefore a linear relationship between two variables will give a correlation coefficient of +1 or —1
                                  depending upon whether large values of one occur with large values of the other or large values of
                                  one occur with small values of the other.
                                  The converse of the above proposition is likewise true, i.e., if the coefficient of correlation (r) equals 1
                                  then the relationship between the X and Y series is linear.
                                  Assume              r = l
                                  then  (   )∑  2  ∑xy  x 2.  ∑–  y 2  = 0

                                                 x
                                        x
                                                            x
                                  Letting 1  =  λ y , 2  = λ y  . . . n  = λ y  the above expression becomes
                                                                 nn
                                                      22
                                             11
                                                                      2
                                    2
                                  yy  2 2  ( 1  1  – λ  2 )λ  2 +y  2  3 2  ( 1  1  λ –  3 )λ y  2   + . . . + yy s 2  ( r  r  λ s )λ –  2  + . . . = 0
                                  The only way in which this expression can equal zero is by having
                                                     λ 1 = λ  = λ  = . . . = λ n
                                                          2
                                                              3
                                  and it follows that
                                                                         x
                                                              x
                                                     x 1 =  λ y , 2  = λ y  . . . n  =  λ y
                                                                              1n
                                                                   12
                                                          11
                                  or
                                                      x = λ y
                                                          1
                                  which denotes a linear relationship between X and Y.
                                  That any relation other than a linear one will not lead to r = l is illustrated by the following:
                                  Let       Y =  X 2
                                            X = 1, 2, 3, 4, 5,              M 1 = 3
                                            Y = 1, 4, 9, 16, 25,           M 2 = 11






         134                              LOVELY PROFESSIONAL UNIVERSITY
   135   136   137   138   139   140   141   142   143   144   145