Ocenjevanje standardne napake in intervalov zaupanja z metodo bootstrap

GAŠPARAC, GRETA (author), Štrumbelj, Erik (mentor)

V diplomskem delu predstavimo metodo bootstrap, ki jo uvrščamo v družino metod samovzorčenja ter je preprostejša in bolj intuitivna alternativa tradicionalnim statističnim metodam za ocenjevanje negotovosti. Osredotočimo se na neparametrično različico metode. Opišemo njene lastnosti in jih predstavimo s praktičnimi primeri s področja strojnega učenja - ocenjevanje in primerjava različnih modelov. Izpostavimo tudi šibke točke metode. Predstavimo in primerjamo tri intervale zaupanja bootstrap: standardnega normalnega z uporabo standardne napake bootstrap in dva klasična intervala bootstrap, centilnega in BCa. Pričakovano se v večini primerov najbolje obnese BCa.

• Language: Slovenian
• Keywords: metoda bootstrap, standardna napaka, intervali zaupanja
• Work type: Bachelor thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2018
• PID: 20.500.12556/RUL-103206
• Publication date in RUL: 14.09.2018

GAŠPARAC, GRETA, 2018, Ocenjevanje standardne napake in intervalov zaupanja z metodo bootstrap [online]. Bachelor’s thesis. [Accessed 14 April 2025]. Retrieved from: https://repozitorij.uni-lj.si/IzpisGradiva.php?lang=eng&id=103206

Secondary language

• Language: English
• Title: Boostraping standard errors and confidence intervals

Abstract

We introduce the reader to the bootstrap, a simple and flexible resampling-based alternative for quantifying uncertainty. We describe the basic characteristics of the non-parametric bootstrap and illustrate its practical behaviour with simulations in the context of a typical task in machine learning - estimating and comparing the performance of different prediction models. We also present some of the method's weaknesses. We introduce and compare three standard intervals: the standard normal using bootstrap standard error and two more typical bootstrap confidence intervals, the percentile and the BCa interval. As theory suggests, the BCa performs the best over a wide range of situations.

• Keywords: bootstrap, standard error, confidence intervals