An integration of cognitive theory and psychometrics: Analogical reasoning.
Item
-
Title
-
An integration of cognitive theory and psychometrics: Analogical reasoning.
-
Identifier
-
AAI9510653
-
identifier
-
9510653
-
Creator
-
Diones, Ruth Ellen.
-
Contributor
-
Adviser: Roger Millsap
-
Date
-
1994
-
Language
-
English
-
Publisher
-
City University of New York.
-
Subject
-
Education, Tests and Measurements | Education, Educational Psychology
-
Abstract
-
The purpose of this dissertation was to explain the difficulty of SAT Verbal analogy items with cognitively relevant, model-based variables by using a componential item response model, the Linear Logistic Test Model (LLTM; Embretson, 1985; Fischer, 1973; Sheehan & Mislevy, 1990; Spada, 1978) in two phases: exploratory and confirmatory. Six main variables were selected from the cognitive literature, based on a model of analogical reasoning, for a priori empirical operationalization. There were three semantic memory variables: written vocabulary knowledge, common semantic relations, and a dichotomy of relations (intensional or pragmatic), and three working memory variables: rationale difficulty, alternative choice influencing decision-making and stem contextualization, and structure-mapping. These variables reflected the following different aspects of a prototypical examinee solving analogy items: (1) the knower, or declarative and procedural memory, (2) the processor, or the utilization of information processes, strategies, or induction/inference within working memory, and (3) the experiencer, or episodic memory.;The data for this study originated from three forms (November 1988, 1989 and 1990) of the SAT used by the Educational Testing Service for equating purposes. Three consecutive forms were necessary in order to provide a large enough set of analogy items: 80 items. The sample of 30,907 was selected from among nondisabled college bound seniors for whom English was the first language with a representative proportion of males and females and ethnicities. The study proceeded in several steps: the cognitive variables were first operationalized and the LLTM model assumptions were checked. Next, an exploratory phase was run on a random subsample in order to test out the LLTM cognitive variables and additionally, to ascertain the importance of the Educational Testing Service's test development taxonomies. Lastly, a final LLTM model was fit on the remaining sub-sample of examinees in a confirmatory phase, thereby assessing the generalizability of the results.;Several variables were important in explaining item difficulty: two semantic memory variables, common Semantic Relations and the key vocabulary frequency; two working memory variables, Level of Relation for structure-mapping and Stem Rationale Difficulty; and one taxonomic variable, Level of Abstraction. Implications for the information processing model, construct validity and test design were discussed.
-
Type
-
dissertation
-
Source
-
PQT Legacy CUNY.xlsx
-
degree
-
Ph.D.
