Construct, which are synonymous with concepts, are theoretical constructions, abstractions, aimed at organizing and making sense of our environment. In other words, construct "is not a visual image, nor is it external to the mind; it is analogous to a piece in a game which thought plays"( Caws, 1959, p.16). Examples of constructs are ansiety, movitivation, mental ability, attitude, self-esteem, interest, frustration, and altruism.

Construct validation is concerned with validity of inferences about unobserved variables(the constructs) on the basis of observed variables (their presumed indicators).  The reciprocity between constructs and indicators was enunciated by Immanuel Kant:"Concepts without factual content are empty; sense data without concepts are blind" (quoted by Mackay, 1977, p.84). Construct validation is fraught with difficulties, ambiguities, even circularities. How, one might well wonder, is it possible to ascertain whether a given observed varoable is an indicator of a variable that is unobservable in the first place? Further, in view of the fact that a given observed variable may reflect different constructs (e.g., the same behavior may reflect different motives), and the same construct may be manifested by different observed variables (e.g., the same behavior may reflect motives), and the same construct may be manifested by different observed variables (e.g., the same motive may be reflected by different behaviors), how does on tell which is which?

Illustrating the inherent ambiguity in making inferences from an indicator to a construct is the following news item:

We hear of a museum in a certain Eastern city that was proud of its amzing attendance record. Recently a little stone building was erected nearby.  Nest year attendance at the museum mysteriously fell off by 100.000. What was the little stone building? A comfort station. (This week, April 17, 1948. Quoted by Waillis & Roberts, 1956, p.133)

Complicating matters further is the fact that, in given situations, the interest is in studying an observed variable in its own right, whereas in others it is treated as an indicator of wome construct. Thus, voting behavior, say, may be studied in its own right, or it may be used as an indicator of some construct, say, political involvement. Or, in a given study, the interest may be in the effect of education on income, whereas in another study both these variables may be taken as indicators of, say, socioeconmoic status.

When observed variables are used as indicators of a construct, care should be exercised not to invest them with the meaning attributed to the construct, as this might lead to erroneous, even ludicrous conclusions. For example, in a reanalysis of data from the influential study of Equality of educational opportunity (Coleman et al., 1966, commonly referred to as the Coleman Report), Armor(1972) found correlations of about.7 between an indes of ownership of nine household items (e.g., television set, vacuum cleaner, telephone, dictionary, refrigerator) and students' verbal achievement. Now, it may be meaningful to regard ownership of the aforementioned items as indicators of a construct (e.g., "family style," "economic well-being of a family", Armor, p.206) and to interpret the correlations accordingly. It is obviously quite a different matter to invest such indicators with the meaning of the construct they presumably represent, as this would lead to the conclusion that it is the owning of a telephone, a refrigerator, a vacuum clener, a refigerator, a vacuum cleaner, and the like that affects students' verbal achievement.

This example may appear so obvious as to not warrant mentioning. Yet misconceptions of this kind are all too common. For examply, reporting on findings of a set of international studies of educational achievement (the IEA studies; see, for example, Peker, 1975), Hechinger(1973)stated:"The number of books and magazines in a student's home had a greater effect on achievement in literature than income and education of the student's family." Without going far afield, it will be noted that it is conceivable for number of books and magazines in the home to be deemed indicators of parents' education or income, not to mention the possiblility of all of the preceding being indicators of , say, socioeconomic status.

        ----------------From Measurement, design, and analysis: an integrated approach. (Elazar J. Pedhazur, Liora Pedhazur Schmelkin,1991, Lawrence Erlbaum Associates, Publishers

转载本文请联系原作者获取授权，同时请注明本文来自王芳科学网博客。
链接地址：https://blog.sciencenet.cn/blog-38036-213990.html

上一篇：My Yellow River
下一篇：Reading notes(2): The Correlation Model: a Comment