Theory of Attitude Measurement

Method of Similar Attributes/Reactions

Suppose a researcher has I statements regarding the attitude toward a particular object (e.g., capital punishment). The first step in the method of similar attributes is to rank order the statements from least favorable to most favorable regarding the object of interest. The method of equal-appearing intervals (also described by Thurstone) might be used to rank order the statements. Next, the coefficient of similarity (denoted Φ) is computed between all pairs of statements using raw data collected from a sample of participants who were asked to endorse statements he or she agrees with. The Φ coefficient represents the degree to which the endorsement of two statements reflects the same attitude. For any statement i and statement j, the Φ-coefficient of similarity is computed as follows:

where nij represents the number of participants who endorsed both statement i and j; ni and nj represent the number of participants who endorsed [Page 1503]statement i and statement j, respectively; and pi and pj represent the reliability of statement i and statement j, respectively. The reliability of a statement represents the probability a participant will endorse the particular statement given he or she has that opinion. In computing the reliability, the fact that the statements have been rank ordered such that any two adjacent statements will reflect practically the same attitude is exploited. Therefore, the reliability of statement i, where statement j is adjacent to statement i, is computed as follows:

Essentially, the reliability of an item represents the proportion of participants who endorsed an adjacent statement that represents the same attitude.

To illustrate the procedure, Table 1 reports hypothetical raw data corresponding to five statements that were selected from a scale consisting of 50 statements administered to 1,000 participants (note that five statements have been selected for illustrative purposes only—i.e., a scale should contain many more items). The diagonal elements in Table 1, which are bolded, report the number of participants who endorsed each statement. The off-diagonal elements indicate the number of participants who endorsed both respective statements. The reliability of each statement is reported in the last row labeled pi and is based on Equation 2, which was applied to the complete scale of 50 statements. The statements in Table 1 have been ordered from least favorable (s1) to most favorable (s5) based on the method of equal-appearing intervals, which was applied to all 50 statements.

...