Testing the Teacher's Report Form Syndromes in 20 Societies.

Abstract. Standardized assessment instruments developed in one society are often used in other societies. However, it is important to determine empirically how assessment instruments developed in one society function in others. The present study tested the fit of the Teacher's Report Form syndrome structures in 20 diverse societies using data for 30,030 6- to 15-year-old students from Asia; Australia; the Caribbean; eastern, western, southern, and northern Europe; and the Middle East. A correlated seven-syndrome model and a hierarchical Attention Problems model were tested separately in each of the 20 societies via confirmatory factor analyses. The results supported the fit of the models in the tested societies. ********** Standardized assessment instruments developed for children in one society are often used in other societies where the development of indigenous instruments may not be practical. However, before assessment instruments developed in one society can be applied in another society, it is important to determine empirically whether they measure the same constructs in both societies. Translations of standardized assessment instruments offer valuable opportunities for testing the applicability of these instruments in different societies. A key step in demonstrating the applicability of an assessment instrument involves testing whether patterns of co-occurring problems identified by the instrument in one society fit the patterns identified by the instrument in the other society. Such patterns of problems can be thought of as "syndromes" (i.e., sets of problems that tend to co-occur). To test the degree to which syndromes found in one society fit data from another society, multivariate analyses are performed on data obtained with the instrument for a large sample of children in the new society. If the syndrome structure derived in the instrument's society of origin fits the data obtained in the new society, the original syndrome structure can be applied in the new society. It is important to note that the obtained syndromes may also be shaped by methodological characteristics of the study, including specific analytic methods, recruitment procedures, sample characteristics, and vicissitudes of translations. The degree to which an instrument's syndrome structure fits a new data set is termed configural invariance. Configural invariance is the most basic component of measurement invariance. Measurement invariance refers to the notion that an assessment instrument measures the same psychological constructs in different populations. In addition to including configural invariance, measurement invariance includes metric invariance (i.e., invariance of factor loadings), scalar invariance (i.e., invariance of intercepts), item residual invariance (i.e., invariance of item error variances), factor variance invariance (i.e., invariance of factor variances), factor covariance invariance (i.e., invariance of factor covariances), and factor mean invariance (i.e., invariance of factor means; Vandenberg & Lance, 2000). The components of measurement invariance can be conceptualized as a pyramid, with configural invariance as the base on which the other components rest. Testing configural invariance is the first step in establishing an instrument's measurement invariance (Vandenberg & Lance, 2000). Factor-Analytic Methods for Deriving and Testing Syndromes Factor-analytic methods are used to derive syndromes. The first stages usually-involve exploratory factor analysis (EFA). EFA is applied to correlations among ratings of items to find patterns of problems that tend to co-occur. After EFA has identified patterns, the patterns can be tested via confirmatory factor analysis (CFA). CFA tests the degree to which a particular model for associations among problems (such as syndromes derived by EFA or by other means) fits a particular data set. If CFA shows good fit between a particular syndrome model and a data set different from the set on which the model was derived, the syndrome model is concluded to have configural invariance in the new data set. …