Modelagem hierárquica de regressão na análise de radiação não ionizante de extrema baixa frequência em residências verticais

• Ano de defesa: 2019
• Autor(a) principal: Silva, Andrezza Araújo Rodrigues da
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal da Paraíba; Brasil; Engenharia de Produção; Programa de Pós-Graduação em Engenharia de Produção; UFPB
• Palavras-chave em Português: Modelo Hierárquico de Regressão; Campo Magnético de Extrema Baixa Frequência; Estação de Transformadores Interna; Hierarchical Regression Model; Extremely Low Frequency Magnetic Field; Internal Transformer Station; CNPQ::ENGENHARIAS::ENGENHARIA DE PRODUCAO
• Link de acesso: https://repositorio.ufpb.br/jspui/handle/123456789/19321

Introduction: Electromagnetic fields have occurred in nature since always, but in the last decades, exposure produced by artificial sources has increased considerably due to the growth in demand for electricity. Although electronic devices, in general, emit high levels of magnetic fields, they use to have their use restricted to short periods of time. Therefore, it is necessary to take into account other significant sources of extremely low-frequency magnetic field (ELF MF), such as the internal transformer stations, present in vertical buildings. Objective: To construct hierarchical regression models for analysis of extremely low frequency non-ionizing radiation in vertical edifications. Methodology: We evaluated 37 environments of 17 apartments located in 14 vertical buildings, in which intensity of magnetic field was measured at the frequency of 60 Hz. Other variables associated with the environment, the apartment and the building were also measured. From these data, multilevel regression models involving four hierarchical levels were constructed, wich are: "measurements" of the intensity of ELF MF (1st level), which were performed with a magnetic spectrum analyzer device; in the rooms (2nd level), which can be of the type bedroom, kitchen, living room, living room-kitchen; which are grouped in "apartments" (3rd level); which constitute each building (4th level). Results and Discussion: (1) Models with random errors at two hierarchical levels. Among the three hierarchical regression models selected, containing the random error term in two levels (nº 1, nº 2 and nº3 transformed), the model nº3, which presents random error in the measurement and environment levels, had the best fit for the AIC and the logarithm of the likelihood function (logLik). For this model, regardless of the variables, the expected mean intensity of ELF MF at 60 Hz is 1.11 T. This value exceeds by 0.71 T the value of ELF MF of 0.4 T, above which Longyo et al. (2014) found a significant statistical association with childhood leukemia. (2) Models with random errors at three hierarchical levels. For the same adjustment criteria, among the three hierarchical regression models selected, containing the random error term in three levels (nº 4, nº 5 and nº 6 transformed), model nº 5, which presents random error in the measurement, environment and apartment levels, had the best fit. In this model, regardless of the fixed effect of the variables, the expected mean intensity of ELF MF at 60 Hz is 1.1294 T. The remaining adjacency variable in the model indicates that apartments adjacent to internal transformers tend to have ELF MF values 39.10% larger than nonadjacent apartments. Regarding the random effects that are not explained by the model variables, 78.04% are explained by the performed measurements , 1.83% by the ambient level and 20.13% by the apartment level. Conclusion: Model nº 5 presented the best fit when comparing the selected models nº 3 and nº 5 transformed according to the AIC and logLik, with the model nº 7 transformed (the only one that was obtained with error in the four hierarchical levels). This model shows that 20.13% of the random effects are due to factors associated with apartments that are not being taken into account by the model, demonstrating that residents of dwellings with internal transformers stations are constantly exposed to levels of extremely low frequency magnetic fields above 0.4 T. Therefore, they are more likely to develop short- and long-term negative health effects.