The hierarchical linear models or multilevels were developed for analysis of data which possess group structure, that is, a structure hierarchy which takes in account the data variability inside and among each hierarchical level. By using data analysis from SAEPE (2002 Educational Evaluation System of Pernambuco), hierarchical models (MH) are presented with two levels of evaluation in mathematics and Portuguese language classes applied to the 4th and 8th grades students of fundamental teaching and to the students 3rd grade students medium teaching. The results in this modeling are more appropriate due to data group structure. A comparison between multiple regression models and hierarchical models shows a better performance of the second model.
The CAUSE Research Group is supported in part by a member initiative grant from the American Statistical Association’s Section on Statistics and Data Science Education