통계 모델링 기법의 비교연구
노유정 · 대한기계학회 춘추학술대회 · 2014
Statistical modeling of input random variables in mechanical systems is necessary to quantify uncertainties arising from inherent variability in material properties and environment variation for accurate analysis of mechanical system behavior. The statistical modeling methods select the best fitted distribution to given data by comparing given data with certain probability density functions or cumulative distribution functions. However, the statistical modeling methods might select different models because they have different measures for the model selection. One method can easily select a right model for skewed data, whereas another method cannot. According to the data distribution or number of samples, each statistical modeling method has different performances, and thus, it is hard for engineers to decide which model is a right model for the given data. This research tests various statistical modeling methods such as Kolmogorov?Smirnov(K-S) test, Chi-square test, AIC(Akaike information criterion), and BIC(Bayesian information criterion) for various types of data. It aims to propose appropriate statistical modeling methods well fitted to data distribution characteristics.
Cited by 0 works.