152 CHAPTER 9 Nonparametric Methods
ρ=− 16 Tnn/[ (^2 − 1 )},
where TRXRY=∑in= 1 [( ) ()]ii−^2.
As an example, let us look at the correlation between the tempera-
ture in Washington and New York over the 12 months of the year. The
reason the paired t test worked so well was because most of the variation
was due to seasonal effects that were removed through the paired differ-
ences. This variation will translate into high correlation between X i , the
temperature in New York on the 15th of month i , with Y i , the temperature
Washington, DC on the 15th of the i th month. Using the ranks, we show
in Table 9.4 how the Spearman correlation is calculated in this case.
Table 9.4
Daily Temperatures for Two Cities: Spearman Rank Correlation
Day Washington mean
temperature ( ° F)
New York mean
temperature ( ° F)
Ranked
pairs
Term
Y (rank) X (rank) [ r ( x ),
ry )]
[ r ( y i ) − r ( x i )]^2
- January 15 31 (2) 28 (3) (3, 2) 1
- February 15 35 (4) 33 (4) (4. 4) 0
- March 15 40 (5) 37 (5) (5, 5) 0
- April 15 52 (6) 45 (6) (6, 6) 0
- May 15 70 (8) 68 (8) (8, 8) 0
- June 15 76 (10) 74 (10) (10, 10) 0
- July 15 93 (12) 89 (12) (12. 12) 0
- August 15 90 (11) 85 (11) (11, 11) 0
- September 15 74 (9) 69 (9) (9, 9) 0
- October 15 55 (7) 51 (7) (7, 7) 0
- November 15 32 (3) 27 (2) (2, 3) 1
- December 15 26 (1) 24 (1) (1.1) 0
T 2
ρ = 1 − 6 T /
[ n ( n 2 − 1)]
= 1 − 1 2 /
[12(143)]
= 142/143
= 0.9930