Background: Research shows that in addition to biological factors, psychological and social risk factors play a role in the development of intellectual disability.
Objectives: This study aimed to investigate the familial, personal and educational factors affecting the intelligence quotient (IQ) levels of intellectually disabled children and adolescents via regularized regression approaches (RRAs) and to compare the results with those of conventional regression approach (CRA).
Methods: A total of sixty characteristics were examined along with dummy variables of 205 children and adolescents selected according to the study protocol. Compact, Lasso, Ridge and Ridged Lasso RRAs were used in the dataset.
Results: The optimal model was obtained with the Lasso approach and contained ten risk factors having a significant effect on IQ levels: diagnosis of cerebral palsy, age at onset of speech, duration of education, age at onset of walking, presence of elimination disorders, presence of attention-deficit hyperactivity disorder, family income, and number of siblings, residence and age. The RRAs provided opportunity to examine more factors than CRAs without requiring the fulfillment of strict CRA assumptions.
Conclusions: Due to the advantages of RRAs, expanding their clinical usage for very large datasets was recommended. [Ethiop. J. Health Dev. 2018;32(1):60-67]
Keywords: Intelligence quotient, intellectual disability, regularized regression, lasso, ridged lasso