Introductory Biostatistics

(Chris Devlin) #1

A 95% confidence intervalðrl;ruÞfor the coe‰cient of correlation under inves-
tigation is obtained by transforming the two endpoints,


zl¼z 1 : 96

ffiffiffiffiffiffiffiffiffiffiffi
1
n 3

r

and


zu¼zþ 1 : 96

ffiffiffiffiffiffiffiffiffiffiffi
1
n 3

r

as follows to obtain the lower endpoint,


rl¼
expð 2 zlÞ 1
expð 2 zlÞþ 1

and the upper endpoint of the confidence interval for the population coe‰cient
of correlation,


ru¼

expð 2 zuÞ 1
expð 2 zuÞþ 1

(In these formulas, ‘‘exp’’ is the exponentiation, or antinatural log, operation.)


Example 4.13 The data shown in Table 4.10 represent systolic blood pressure
readings on 15 women. The descriptive analysis in Example 2.9 yieldsr¼ 0 : 566
and we have



1


2


ln

1 þ 0 : 566
1  0 : 566
¼ 0 : 642

zl¼ 0 : 642  1 : 96

ffiffiffiffiffi
1
12

r

¼ 0 : 076


TABLE 4.10


Age (x) SBP (y) Age (x) SBP (y)


42 130 80 156
46 115 74 162
42 148 70 151
71 100 80 156
80 156 41 125
74 162 61 150
70 151 75 165
80 156

ESTIMATION OF CORRELATION COEFFICIENTS 169
Free download pdf