Testing the Invariance of Reading Literacy Dimensions across Different Countries: An Application of Multigroup Confirmatory Factor Analysis.

This study investigates whether or not the factor structure of reading comprehension is invariant across large, nationally representative samples of 14-year-old students from four different countries. The data from French-speaking Belgium, Hungary, Italy, and the United States were collected as part of the Reading Literacy Study of 1990-91, conducted by the International Association for the Evaluation of Educational Achievement. The relevance and application of multigroup confirmatory factor analysis techniques (K. G. Joreskog and D. Sorbom, 1988) to the assessment of model generalizability across countries or cultures, particularly in relation to international databases, is demonstrated and discussed. Results indicate that it is not unreasonable to assume factor loadings and factor correlations to be invariant across the four countries. When item uniquenesses were set to be invariant, a decrease in fit according to the relative noncentrality index and the Tucker Lewis index was observed. In contrast, the parsimony relative noncentrality index showed a higher value for the model in which invariance was assumed for all parameters in the model. (Contains 1 table, 1 figure, and 27 references.) (Author/SLD)