A1320
Title: A robust longitudinal study of the influence of air pollutants on children
Authors: Ian Danilevicz - CentraleSupelec (France)
Pascal Bondon - CentraleSupelec (France) [presenting]
Valderio Anselmo Reisen - DEST-CCE-UFES (Brazil)
Faradiba Sarquis - Federal University of Espirito Santo (Brazil)
Abstract: To measure the influence of environmental predictors on allergic children, we combine a robust principal component analysis (RPCA) and a robust estimation of a linear mixed model (LMM) in a longitudinal study. This strategy is justified by the verification that robust estimation of a LMM might not deal with correlated outliers in covariates. RPCA transforms the original covariates into uncorrelated variables which are used as new covariates in the LMM. In the real data analysis, RPCA exhibits three principal components mainly related with humidity and particulates matter with a diameter smaller than 10 micrometers (PM10) and 2.5 micrometers (PM2.5), respectively. The study shows that high levels of total immunoglobulin E, dry weather and PM10 are significant risk factors for children's respiratory diseases.
