| Summary: | This thesis study about goodness-of-fit testing approach for normality based on
Bayesian method and its comparisons with classical method. In this context, we
mainly focus on the normality test because it is an important assumption in many
statistical methods. Classical method is normality test which can be regarded as
commonly method, such as Shapiro-Wilk test, Anderson-Darling test, and
Cramer-von Mises test. Bayesian method is used for determining posterior
predictive distribution to obtain the predictive sample. Monte Carlo simulation is
carried out to evaluate the comparison between classical method and Bayesian
method. Alternatives distributions that is considered in the simulation are
symmetric long-tailed distribution, asymmetric long-tailed distribution, mixednormal
distribution, and short-tailed distribution. This paper shows that Bayesian
method is more powerful than classical method against a variety of alternative.
|
|---|