We establish a principle component analysis in the context of two-step monotone incomplete data drawn from a multivariate normal population. Data consists of n observations of p+q variables and additional N-n observations of p variables, where all observations are mutually independent. We perform the principal component analysis using the maximum likelihood estimator with the monotone incomplete data, derive the asymptotic distribution of the latent root of covariance matrix, and compare them with the results of the typical canonical correlation.

Keywords: principle component analysis; two-step monotone incompete data; asymptotic distribution of latent root of covariance matrix

Biography: I am an associate professor at the Faculty of Economics, Sapporo Gakuin University. My research area is the multivariate analysis, missing problem and asymptotic theory.