Uma abordagem estatística sobre a estimação de redshifts de quasares usando dados do S-PLUS

• Autor(a) principal: Soares, Gabriela Pereira
• Data de Publicação: 2022
• Tipo de documento: Trabalho de conclusão de curso
• Idioma: por
• Título da fonte: Repositório Institucional da UFSCAR
• Texto Completo: https://repositorio.ufscar.br/handle/ufscar/16618
• topic: Densidade condicional; Estimação; Filtro de banda estreita; Quasar; FlexCode; Redshift; S-PLUS; CIENCIAS EXATAS E DA TERRA::ASTRONOMIA; CIENCIAS EXATAS E DA TERRA::PROBABILIDADE E ESTATISTICA

Redshift is a cosmic index used to measure distances to astronomical objects. The study of this quantity is important for the understanding of the expansion of the Universe and the current objective of the stars, according to cosmology. In this work, we are interested in estimating distances of quasars, which are luminous celestial objects known by its high redshifts, indicating that they are at great distances from Earth. The estimation of redshift can be performed via spectroscopy, but this technique has a high cost and requires a large amount of time for cosmic observation. Thus, photometric surveys have been highly valuable in this field, as they also provide relevant information for measuring redshift, despite having low resolution and less precision. The goal of this work is to improve the estimation of photometric redshifts for quasars from S-PLUS (Southern Photometric Local Universe Survey). In order to do that, we build statistical models based on the estimation of conditional densities using the FlexCoDE algorithm. In addition, we study the influence of narrowband filters (narrow bands) on the model, currently available only in S-PLUS, and compare it with the results of a previously developed neural network model, with the purpose of confirming the significance of these bands. We found from the analysis that narrow bands significantly improve the estimates of the conditional density of the photometric redshift, although this improvement is not observed in point estimators for the redshift. However, the diagnosis detected that the tested models, both from FlexCoDE and from neural networks, may be improved.