Empirical Likelihood Method for a Location Parameter Using Some Robust Estimators
Loading...
Date
Authors
Advisor
Journal Title
Journal ISSN
Volume Title
Publisher
Latvijas Universitāte
Language
eng
Abstract
Pētījumā attīstītas empīriskās ticamības (EL) metodes divu un vairāku neatkarīgu populāciju salīdzināšanai, balstoties uz robustiem lokācijas parametra novērtētājiem. Iegūti jauni asimptotiskie rezultāti par empīriskās ticamības metodēm: 1) divu M-novērtētāju starpībai; 2) divu nošķeltu vidējo vērtību starpībai; 3) uz empīrisko ticamību balstītai ANOVA metodei vairāk kā divu nošķeltu vidējo vērtību salīdzināšanai. Tika izstrādāts simulāciju eksperiments un analizēti datu piemēri, kas parādīja, ka jauniegūtās metodes ir līdzvērtīga alternatīva klasiskās statistikas metodēm gadījumos, kad dati ir normāli sadalīti – tām ir līdzīga jauda un spēja kontrolēt empīrisko pirmā veida kļūdu. Turklāt metodēm ir labas robustuma īpašības, pārspējot klasiskās metodes gadījumos, kad normalitātes pieņēmums neizpildās. Atslēgas vārdi: empīriskā ticamība; gludinātie M-novērtētāji; robustā statistika; hipotēžu testi; divu izlašu problēma; ANOVA
In this research empirical likelihood (EL) methods for comparing two and multiple independent populations based on robust location estimators are developed. New asymptotic results are proven for the following empirical likelihood-based methods. 1. The difference of two M-estimators, 2. the difference of two trimmed means and 3. EL-based ANOVA method for comparing multiple trimmed means. A simulation study was designed and data examples were analysed showing that the newly-established methods provide a comparable alternative to the classical procedures when the data is normally distributed, demonstrating similar power and ability to control the type I error. In addition, the methods have good robustness properties, having an advantage over the classical procedures when the assumption of normality does not hold. Keywords: empirical likelihood; robust statistics; M-estimator; smoothed M-estimator; trimmed mean; two-sample problem; EL ANOVA
In this research empirical likelihood (EL) methods for comparing two and multiple independent populations based on robust location estimators are developed. New asymptotic results are proven for the following empirical likelihood-based methods. 1. The difference of two M-estimators, 2. the difference of two trimmed means and 3. EL-based ANOVA method for comparing multiple trimmed means. A simulation study was designed and data examples were analysed showing that the newly-established methods provide a comparable alternative to the classical procedures when the data is normally distributed, demonstrating similar power and ability to control the type I error. In addition, the methods have good robustness properties, having an advantage over the classical procedures when the assumption of normality does not hold. Keywords: empirical likelihood; robust statistics; M-estimator; smoothed M-estimator; trimmed mean; two-sample problem; EL ANOVA