Statistics for the Behavioral Sciences by Frederick J. Gravetter, Larry B. Wallnau ISBN 10: 1305504917 ISBN 13: 9781305504912

516 CHAPTER 15 | Correlation
that a sample correlation should be representative of the corresponding population value.
In particular, if the population correlation is ρ S
= 0 (as specified in H 0
), then the sample
correlation should be near zero. For each sample size and alpha level, the table identifies
the minimum sample correlation that is significantly different from zero. The following
example demonstrates the use of the table.
EXAMPLE 15.14
An industrial psychologist selects a sample of n = 15 employees. These employees are
ranked in order of work productivity by their manager. They also are ranked by a peer. The
Spearman correlation computed for these data revealed a correlation of r S
= .60. Using
Table B.7 with n = 15 and α = .05, a correlation of ±.521 is needed to reject H 0
. The
observed correlation for the sample easily surpasses this critical value. The correlation
between manager and peer ratings is statistically significant.
■
It is customary to use
the numerical values 0
and 1, but any two different
numbers would
work equally well and
would not affect the
value of the correlation.
■ The Point-Biserial Correlation and Measuring Effect Size with r 2
In Chapters 9, 10, and 11 we introduced r 2 as a measure of effect size that often accompanies
a hypothesis test using the t statistic. The r 2 used to measure effect size and the
r used to measure a correlation are directly related, and we now have an opportunity to
demonstrate the relationship. Specifically, we compare the independent-measures t test
(Chapter 10) and a special version of the Pearson correlation known as the point-biserial
correlation.
The point-biserial correlation is used to measure the relationship between two variables
in situations in which one variable consists of regular, numerical scores, but
the second variable has only two values. A variable with only two values is called a
dichotomous variable or a binomial variable. Some examples of dichotomous variables
are:
1. male vs. female
2. college graduate vs. not a college graduate
3. first-born child vs. later-born child
4. success vs. failure on a particular task
5. older than 30 years old vs. younger than 30 years old
To compute the point-biserial correlation, the dichotomous variable is first converted
to numerical values by assigning a value of zero (0) to one category and a value of one
(1) to the other category. Then the regular Pearson correlation formula is used with the
converted data.
To demonstrate the point-biserial correlation and its association with the r 2 measure
of effect size, we use the data from Example 10.2 (p. 310). The original example compared
cheating behavior in a dimly lit room compared to a well-lit room. The results
showed that participants in the dimly lit room claimed to have solved significantly
more puzzles than the participants in the well-lit room. The data from the independentmeasures
study are presented on the left side of Table 15.4. Notice that the data consist
of two separate samples and the independent-measures t was used to determine whether
there was a significant mean difference between the two populations represented by the
samples.
On the right-hand side of Table 15.4 we have reorganized the data into a form that is
suitable for a point-biserial correlation. Specifically, we used each participant’s puzzlesolving
score as the X value and we have created a new variable, Y, to represent the group
or condition for each individual. In this case, we have used Y = 0 for individuals in the
well-lit room and Y = 1 for participants in the dimly lit room.