Chapter 7 Tables 303inherent order. For ordinal variables, there are more powerful tests than
the Pearson chi square, which often fails to give signifi cance for ordered
variables.
As an example, consider the Calculus and Enroll B variables. Since the
Calculus variable tells the extent to which calculus is required for a given
statistics class (Not req or Prereq), it can be treated as either an ordinal or
a nominal variable. When a variable takes on only two values, there really
is no distinction between nominal and ordinal, because any two values can
be regarded as ordered. The other variable, Enroll B, is a categorical vari-
able that contains measures of the size of the annual course enrollment. In
the survey, instructors were asked to check one of eight categories (0–50,
51–100, 101–150, 151–200, 201–300, 301–400, 401–500, and 501–) for the
number of students in the course. You might expect that classes requiring
calculus would have smaller enrollments.Testing for a Relationship between Two Ordinal Variables
We want to test whether there is a relationship between a class requiring
calculus as a prerequisite and the size of the class. Our hypotheses are
H 0 : The pattern of enrollment is the same regardless of a calculus prerequisite.
Ha: The pattern of enrollment is related to a calculus prerequisite.
To test the null hypothesis, fi rst form a two-way table for categorical vari-
ables, Calculus and Enroll B.To form the table:1 Return to the Survey worksheet and create another PivotTable on a
new worksheet.
2 Place the Enroll B fi eld in the Row Labels area of the PivotTable
and Calculus in the Column Labels area. Also place Calculus in the
Values area of the table.
3 Remove blank entries from both the row and column labels in the
PivotTable. Figure 7-24 shows the fi nal PivotTable.