Self-Organizing Maps Applied as Analysis Tool of Reading Cognitive Test

Abstract

This paper proposes the use of Self-Organizing Maps (SOM) as a tool for exploratory data analysis related to the study of cognitive abilities, an important research area on Educational Psychology that, like other areas such Psychology and Behavioral Sciences uses inferential statistics techniques as a way of confirming the validity of researches. These statistical methods are fundamental for researchers, but they also have a well-documented set of limitations, such as issues related to difficulty in choosing the best methods of analysis, difficulty in interpreting the results, problems in hypotheses formulation, problems related to sampling, such as sample size, occurrence of non-random errors and the presence of outliers, which if not properly addressed generate deformations in the results to the point of classic authors in data analysis as John Tukey, Jacob Cohen and later the APA (American Psychology Association), began to recommend the use of exploratory analysis in a complementary way to Inferential Statistics. It is hoped to show that SOM can help educational researchers to provide better diagnoses related to students' cognitive development, helping in the validation and normalization of cognitive tests.