You are in the accessibility menu

Please use this identifier to cite or link to this item: http://acervodigital.unesp.br/handle/10400.2/2071
Full metadata record
DC FieldValueLanguage
dc.contributor.advisorRamos, Maria do Rosário-
dc.contributor.authorFaria, Bruno Fernando Pinheiro-
dc.date.accessioned2012-03-08T14:27:39Z-
dc.date.accessioned2017-12-14T17:39:24Z-
dc.date.available2012-03-08T14:27:39Z-
dc.date.available2017-12-14T17:39:24Z-
dc.date.issued2011-
dc.identifier.citationFaria, Bruno Fernando Pinheiro - Teste F na regressão linear múltipla para dados temporais em correlação serial [Em linha]. Lisboa : [s.n.], 2011. 96 p.por
dc.identifier.urihttp://hdl.handle.net/10400.2/2071-
dc.identifier.urihttp://acervodigital.unesp.br/handle/10400.2/2071-
dc.descriptionDissertação de Mestardo em Estatística, Matemática e Computação apresentada à Universidade Abertapor
dc.description.abstractNa análise de regressão linear, simples ou múltipla, o teste F é utilizado para testar simultaneamente a significância de um conjunto ou um subconjunto de parâmetros. Neste trabalho, é estudado o comportamento da estatística F usual para testar a significância dos coeficientes sazonais no modelo de regressão linear múltipla para séries temporais com tendência, sazonalidade e correlação serial. Quando alguns dos pressupostos de validade do teste são violados é de se esperar que o teste seja afetado. Por isso, analisa-se, através de um estudo de simulação de Monte Carlo, o comportamento da estatística F quando são violados os pressupostos de normalidade e da independência dos erros num caso específico de modelo de autocorrelação – AR(1). O estudo de simulação para avaliar a performance do teste F usual foi realizado sob vários cenários, tendo em conta desvios da normalidade e considerando que os erros do modelo são correlacionados, com diferentes intensidades da autocorrelação. O comportamento dos testes é avaliado com base nos valores do nível de significância empírico e da potência dos testes. Além disso, apresenta-se teoricamente a estatística de teste F para casos em que a autocorrelação nos erros é considerada através da estimação pelos mínimos quadrados generalizados e realiza-se, através de simulação, um estudo comparativo do comportamento deste teste, quando a autocorrelação é incorretamente estimada. Finalmente são apresentadas duas aplicações em séries temporais, uma respeitante ao número de hóspedes em hotéis tradicionais e casas de turismo rural nos Açores e a outra relativa ao número de exportações de automóveis em Portugal.por
dc.description.abstractIn simple or multiple linear regression analysis, the F test is used to test simultaneously the significance of a set or a subset of parameters. In this assignment it is studied the behaviour of the usual F statistic to test the significance of seasonal coefficients in multiple linear regression model for time series with trend, seasonality When some of the test assumptions are violated, we have to assume that the test can be affected. Therefore, we will analyse, through a Monte Carlo simulation study, the behaviour of the F statistic, when the normality assumptions and the independence of the error term in a specific case of autocorrelation model AR(1) are violated. The simulation study to assess the performance of the usual F test was made under several situations, bearing in mind normality deviations and considering that the errors of the autoregressive model are correlated with different autocorrelation levels. The behaviour of the tests is evaluated through the empirical significance level and the power of the test. Furthermore, we present theoretically the F statistic in cases in which the autocorrelation of the error term is taken into account through general least squares estimation and we make, through a simulation study, a comparative analysis of the aforementioned test behaviour, when the autocorrelation of the error term is incorrectly estimated. Finally, we will show two applications in a time series, one regarding the number of guests in traditional hotels and rural tourism in Azores and the other the number of cars exportations in Portugal.-
dc.language.isoporpor
dc.rightsopenAccesspor
dc.subjectMatemáticapor
dc.subjectEstatísticapor
dc.titleTeste F na regressão linear múltipla para dados temporais em correlação serialpor
dc.typeoutropor
Appears in Collections:Dissertações de Mestrado - Universidade Aberta de Portugal

There are no files associated with this item.
 

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.