ERIC Number: ED178599
Record Type: RIE
Publication Date: 1979-Mar
Reference Count: 0
An Introduction to Generalizability Theory as a Contributor to Evaluation Research.
Gillmore, Gerald M.
It is argued in this paper that generalizability theory provides a uniquely useful framework for defining and quantifying the dependability of data for decision making. It does so by requiring careful specification of the conditions of measurement and the anticipated sources of variation in the results of the measurement procedure. A distinction is made between generalizability (G) studies and decision (D) studies. A G study is conducted to determine certain characteristics of the measuring device, while a D study is conducted in support of decision making. The decision-maker must specify (1) the object of measurement; (2) the universe of generalization; (3) the conditions to be sampled from each facet; and (4) the experimental design. Summary statistics for D studies may include: the expected observed score variance, the universe score variance, three types of error variance, and indices of dependability, including the generalizability coefficient. An application of the generalizability coefficient is described and discussed. Confidence intervals are discussed briefly. The testing of persons as individuals, rather than as class members, is discussed and compared for norm referenced and domain referenced tests and specifically for mastery tests. (CTM)
Publication Type: Reports - Research
Education Level: N/A
Authoring Institution: Washington Univ., Seattle. Educational Assessment Center.
Identifiers: Classical Test Theory; Generalizability Theory