Confiabilidad intercalificadores y validez de constructo del test gestáltico de bender (segunda versión)

Abstract

Although the impact of the measurement error in the accuracy of the classification of subjects and validity correlations is theoretically established, in a practical situation has not been explored the degree of impact on the new Bender-II (Brannigan & Decker, 2003). Structural changes con sist ed in more of items (16 designs), comple mentary tests (fine motor and visual perception) and two major tests (Visual Constructive Memory and Visual Motor), the rating method Global Scoring System (GSS) and standardized record sheet for the child's behavior during the test administration. The GSS was created ad hoc for the Bender-II, and it is a method that emphasizes the exact reproduction of the designs; its origin is in the original gestalt approach of Bender. Main studies have been published in the manual (Brannigan & Decker, 2003), and subsequent studies have used the American standardization sample. However, in nonimmigrant Hispanic population, to date there are some unpublished and published only one (Me ri no, 2012); therefore, it is not known how generalizable the findings and psychometric properties obtained in the American standard ization sample are. The aim of the study was to examine the effect of the variability of the scoring of designs of the new Bender-II on construct validitywith a measure of intelligence. The Bender-II and Kaufman Brief Intelligence Test (K-BIT-Kaufman, A.S. & Kaufman, A.L., 1994) was administered to 60 pre-school children (between 4 and 5 years, 33 girls), and three scorers rated the designs reproduced, by standardized procedures and GSS. The analysis consisted of two steps: first, we estimated the consistency and inter-rater agreement using random two-way intraclass correlations (McGraw & Wong, 1996; Shrout & Fleiss, 1979). Second, for each scorer, correlations were calculated between scores on the Bender-II and K-BIT, and finally these correlations were compared with a test for dependent correlations with a common element (Steiger, 1980, Williams, 1959). The results indicate that there were slight differences between two scorers, but one of them had comparatively lower coefficients of consist ency and agreement. In the all scorers, the mag nitude of the consistency coefficients (> .85) and agreement (> .84) between the qualifiers indicate good levels of concordance with even moderate exercise time (two or three sessions). Correlations between scores on the Bender-II and K-BIT were around .43 and .61, and the lowest correlations occurred in the scorer that showed less consistency and agreement with other scorers, clearly indi cating the impact of measurement error in the validity correlations. Compared with one of the rating, this difference was statistically significant, and the percentage reduction of covariance was at least 5%. Finally, the results indicate several points. First, we present other evidence for inter-rater reliability of the Bender-II, and that there were good levels of agreement and consistency. Second, there is potentially a reduction in the correlations of validity when a scorer has trouble for interpret consistently and correctly Global Scoring System. Correlations decreased, however, are not equal in all the scorers, and therefore must be verified the goodness of fit between scorer and the Bender it's qualification method approach. The results can be are idio-syncratic to the sample and study conditions, and the sample size constraints threaten the gene-ralizability of the findings. However, the study conditions are close to professional practice and therefore can be generalized to some extent. In addition, can serve as a baseline to compare future studies of reliability in the Bender-II.