Dating-partner preferences among a group of inner-city African-American high school students

Adolescence, Spring, 1996 by Sherry P. Smith

Variables

Selection of 12 variables for the questionnaire was based mainly on previously cited work: "is caring," "is honest," "has good clothes," "has a nice car," "has a sense of humor," "dances well," and "is good-looking" (Waller, 1937; Hansen & Hicks, 1980; Buss & Barnes, 1986). The variables "is religious" and "is athletic" were selected primarily because of the researcher's interest in how these traits would be perceived by high school students. Two variables were rephrased versions of previously used variables: "makes good grades" instead of "is intelligent," and "is fun to talk to" instead of "personally pleasant" (Waller, 1937; Christensen, 1950; Hansen & Hicks, 1980; Buss & Barnes, 1986). The variable "boring" was chosen solely to determine the occurrence of a response set.

Initial Procedure

In order to accurately represent the data set of 80 responses and 12 variables, a rank-ordering of mean scores was generated by SPSS for the entire group of students; subsequently, rank-orderings were done for males and females. (Since 63% of the sample were females, it was expected that the overall rank-ordering would more closely resemble female responses.)

Due to the nature of the variables, the level of measurement is ordinal; therefore, no "true" value can be assigned to the variable scores. Here, variables are measured in terms of "more" or "less" importance relative to other variables. By assigning measures to these empirical observations, it is possible to make inferences of otherwise unobservable patterns. The values were coded as having a score of "1" for responses of "very important," "2" for responses of "somewhat important," and "3" for responses of "not important." That is, if "caring" receives a "very important" from a respondent, and "boring" receives a "not important" response, then "caring" would receive a "1" and "boring" a "3". The different scores reflect a difference in the level of importance of each variable. Scores closest to "1" indicate that students, on average, rate this variable "very important."

Second Procedure

According to Jacoby (1991), multidimensional scaling (MDS) allows the researcher to represent variables in space so that these relationships reveal the underlying attributes being measured. In order to view how variables relate to each other based on the 80 self-reports, the SPSS program "ALSCAL" is used to obtain a multidimensional scaling of the rated variables. Previous studies implicitly suggest more than one dimension in students' evaluations of potential dating partners, but researchers have virtually neglected to represent possible multidimensionality (e.g., Goodwin, 1990). The general multidimensional scaling model provides an ordinal, monotonic (nonlinear) relationship among the 12 variables. In order to view these relationships, the values can be illustrated on a set of axes. The present data set is input here as one-mode, two-way (a row-by-column matrix of variables).

In multidimensional scaling, a configuration of proximities among the variables enables the observer to more readily recognize patterns in the data that otherwise might not be evident. In the ALSCAL program, the paired relationships are read as dissimilarities whereby smaller numbers represent "closer" entities and larger numbers represent entities which are "far apart." Specifically, ALSCAL transforms proximities into Euclidean distances since a degenerate solution sometimes obtains where correlation coefficients are used. Therefore, if the proximity between "caring" and '"honesty" is smaller than that between "caring" and "good clothes," then the output distance between "caring" and "honesty" should be smaller than the distance between "caring" and "good clothes." Using Young's S-Stress Formula 1, we can observe the "stress" or "badness of fit" of the scale to the data. This formula provides an important criterion for determining which model best conforms to the data. However, we must also consider more practical issues since the complexity inherent in increased dimensions undermines the ability to interpret the output in a comprehensible, meaningful way.