Métodos analíticos de regressão linear com erros em ambas varíaveis

Carregando...
Imagem de Miniatura

Data

2017-06-22

Autores

Malatrasi, Jorge Luis Georges [UNESP]

Título da Revista

ISSN da Revista

Título de Volume

Editor

Universidade Estadual Paulista (Unesp)

Resumo

Currently, in many scientific works, we realize that the statistical analysis are often left aside, or just superficially treated. This work has the aim of showing the importance of the due statistical treatment, with regard to the straight line fitting, using known statistical methods, like least squares method, goodness of fit and others. In particular, we show the importance of considering data uncertainties in both variables in several situations. From this, by analyzing data from papers available in the literature, we applied the methods studied in the present work, compared the results obtained in those papers, showing the difference in the final result, where we have obtained more statistically correct results. As general results, we have found that in case of straight line fitting with fixed uncertainties in both variables, the best fit parameters depend on the uncertainties ratio. We also have found that the best fit depend more of the ratio of uncertainties in the case that the data are poorly correlated
Atualmente em diversos trabalhos, percebemos que a análise estatística muitas vezes é deixada de lado, ou tratada muito superficialmente. Este trabalho tem como objetivo mostrar a importância do devido tratamento estatístico, no que diz respeito ao ajuste de retas, utilizando métodos estatísticos conhecidos, como o método dos mínimos quadrados, qualidade do ajuste e outros. Em particular, mostramos a importância de se considerar as incertezas dos dados em ambas variáveis em várias situações. A partir disso, analisando dados de artigos disponíveis na literatura, aplicamos os métodos estudados no presente trabalho, comparamos com os resultados obtidos naqueles artigos, mostrando a diferença no resultado final, onde obtivemos resultados estatisticamente mais corretos. Como resultados gerais, encontramos que para ajuste de reta com incertezas fixas em ambas variáveis, os melhores ajustes dos parâmetros dependem da razão das incertezas. Encontramos também que os melhores ajustes dependem mais da razão das incertezas no caso em que os dados são pouco correlacionados

Descrição

Palavras-chave

Estatística, Análise de regressão, Estatistica matematica, Statistics

Como citar

MALATRASI, Jorge Luis Georges. Métodos analíticos de regressão linear com erros em ambas varíaveis. 2017. . Trabalho de conclusão de curso (bacharelado - Engenharia Industrial Madeireira) - Universidade Estadual Paulista Júlio de Mesquita Filho, Câmpus Experimental de Itapeva, 2017.