Bootstrap procedure for studies on
material fatigue data

Mariano MARTINEZ-ESPINOSA1

Vera Lúcia Martins SANDANIELO1

Francisco LOUZADA-NETO2

§    ABSTRACT: The term bootstrap is generally used to refer to a simulation technique that aims at obtaining interval confidences for the estimation of parameters of interest by resampling the original data set. The Bootstrap technique can be specifically used when polynomial regression models are considered. A special case of these models is obtained when the values of the independents variables are coded by using a factorial design. In this case, the independent variables are orthogonal, leading to more accurate parameter estimates. The objective of this study is to use both residual and pairs bootstrap procedure on experimental data on fatigue strength in wood specimens, obtained from the Laboratory of Wood and Timber Structures (LaMEM)-EESC-USP-Brazil for checking the adequacy of the normal multiple orthogonal polynomial model in the study of fatigue strength in wood specimens and estimate the parameters in this model at different experimental situations by using simulation. The results of the simulations by the bootstrap procedures show that the bootstrap technique can be used for the objective of this study.

§    KEYWORDS: Bootstrap; fatigue strength; wood; normal multiple orthogonal polynomial model.



1Departamento de Estatística, Instituto de Ciências Exatas e da Terra, Universidade Federal de Mato Grosso ‑ UFMT, CEP 78060-900 Cuiabá, MT, Brasil. E-mail: marianom@cpd.ufmt.br / veluma@cpd.ufmt.br

2Departamento de Estatística, Universidade Federal de São Carlos - UFSCar, Caixa Postal 676, CEP 13565‑905, São Carlos, SP, Brasil. E-mail: dfln@power.ufscar.br