In principal component analysis. usual estimation of principal components is based on data from identically independent and equal distribution. However, if principal components estimate using data that collected by complex sample survey, then assumpt...
In principal component analysis. usual estimation of principal components is based on data from identically independent and equal distribution. However, if principal components estimate using data that collected by complex sample survey, then assumption be not satisfied that identically independently selected sample from equal distribution, because sampling probability differ each other. This thesis will deal with estimation problem of principal components from data collected through complex sample survey. We estimate weighted variance-covariance matrix and weighted correlation coefficient matrix from complex survey data, and study process of estimating principal components using it. After theoretical study, we estimated principal components using Korea Welfare Panel Survey data. In estimation process, we numerically compare ordinary principal components estimator and weighted principal components estimator. As a result, in viewpoint of bias, weighted principal components estimator is smaller than ordinary principal components, but in viewpoint of variance, weighted principal components estimator is larger than ordinary principal components.