Selection of terms in polynomial regression model

Abstract

The problem of determining the "best" subset of variable has long been of interest to statisticians. And in response surface experiments a polynomial model is often used to fit the response surface. In this thesis, we considered how to select subsets of polynomial terms from a given model so as to achieve better fitting surfaces to estimate the response. We introduce an unbiased estimator of the criterion Q and compare a biased estimator of it with the unbiased one for the selection of polynomial terms and illustrated with an example.;이 논문에서는 Regression Model 을 보다 잘 적합시키기 위해, 주어진 Model로 부터 Polynomial항들의 Subset을 선택하는 문제에 대해 다루어진다. 그 방법으로서 criterion Q의 unbiased 추정치를 도입하고, 예를 통해 Polynomial 항들의 선택에 있어 biased Q~과 unbiased Q^을 비교해 본다.