Ansiedade face ao teste e as autocrenças acadêmicas: seu impacto no desempenho em avaliações em larga escala

Abstract

This work consists of two distinct studies with different objectives about the impact of the Test Anxiety and Academic self-beliefs on performance in large-scale assessments. The analyzes used the database of the Programa de Avaliação da Rede Pública de Educação Básica (PROEB) 2011 edition, which evaluated subjects Portuguese and mathematics at the 5th and 9th years of Elementary Schooll and the last year of High School of municipal and state schools of Minas Gerais.The database included 661,011 students and 9,920 schools. Both studies consisted in the improvement of a study conducted in PROEB 2010.In the first study, several changes in the Test Anxiety Scale (AFT), Academic Self-Efficacy Scale (AFA), Academic Self-Concept Scale (ACA) and Perceived Control Results Scale (PC) were performed.Then we analyzed the impact of changes in the quality of the scales. Also tested the suitability of a new structural model configuration consisted by the variables AFT, APA, ACA and MTA and a second-order factor named Academic self-beliefs.The second study assessed the impact of these variables on student performance through a hierarchical linear model, considering three levels: student, class and school. The impact coefficients of the variables of the model were compared with those obtained in research conducted in PROEB 2010.In the first study, to evaluate of the scales, a factor analyzes were conducted in two stages, exploratory and confirmatory. Then, comparisons between Cronbach's alpha of the current scale and the previous one were performed.The method used was Promax rotation. The structural equation modeling allowed that the structure of factors found were tested and compared with the structure of factors found in the study conducted in PROEB in 2010.In the second study, the AFT factor and second-order factor, Academic self-beliefs, had their scores generated by the Item Response Theory, model Samegima, and entered into the hierarchical linear model, along with the other variables in the analysis.The coefficients of hierarchical linear models of 2010 were compared to the coefficients of the present study. The results of the first study, shown the emergence of a new factor, Academic Motivation and the demise of the PC factor.The Cronbach's alpha of the scales was higher than the previous scales, except the AFA factor. Alphas were: AFT: 0.842; AFA: 0.793; ACA: MTA and 0.911: 0.754. Two structural models were produced. One of them used the data structure proposed in the 2010 study, however, the measures of the new scales (this model was called the 2nd model).The other model used the new data structure found in the confirmatory factor analysis and measures of the scales produced for this study (model 3). These models were compared and the 3rd model demonstrated good fit, superior to other models: CFI = 0.96; NFI = 0.96; RMSEA = 0.12 and RMR = 0.05.In the second study, the analisys was made by the comparison of coefficients of hierarchical linear models, considering only the level of the student and was focused only on emotional variables. The coeficients of the variables, AFT and Academic self-beliefs of this study were higher than coefficients to emotional variables in the 2010 model, especially the academic selfbeliefs factor.It is believed that the difference is due to the improvement of scales, which become more specific. The emotional variables had the greatest impact on performance coefficients in the linear model, higher then socioeconomic index and school delay, variables that are traditionally knowed as a varialble of large impact on performance . The consequences of the impact of emotional variables on performance are discussed from the point of psychometric and educational standpoint.