zahra naghsh; azam moghadam
Volume 2, Issue 8 , July 2012, , Pages 133-154
Abstract
In traditional approaches, single-level statistical models were generally used to analyze IEA data. In hierarchical linear models, each level is, however, introduced by its sub-model, and the variables' interrelations are explained in each specified level. The way the variables affect the existent relations ...
Read More
In traditional approaches, single-level statistical models were generally used to analyze IEA data. In hierarchical linear models, each level is, however, introduced by its sub-model, and the variables' interrelations are explained in each specified level. The way the variables affect the existent relations in other levels is also determined. The main purpose of this paper is to compare multi-level modeling and single-level analysis techniques and underline the importance of applying the former in analyzing the data extracted from the TIMSS 2007 questionnaires completed by the eighth graders. Due to their nature, the IEA data were analyzed by HLM software as the students were nested within classes, classes within schools, and schools within countries. In the single-level analysis, there was a significant relationship between self- concept, attitude and evaluation at 0.001 level with mathematics achievement (0.48, 0.296 and 0.134, respectively). Furthermore, the results of two-level analysis by one-way ANOVA with random effects showed that these three variables (self-concept, attitude and evaluation) explained 30.10% and 47% of mathematics achievement variance at student and school levels, respectively. The different results of these two analyses demonstrated the importance of using multi-level analyses for nested data like TIMSS. Regarding the nested nature of TIMSS data and the multi-level method used to extract them, the application of multi-level modeling techniques is recommended to obtain more detailed data on the factors influencing the students' achievement.
