Inferences on a Normal Covariance Matrix and Generalized Variance with Monotone Missing Data

The problems of testing a normal covariance matrix and an interval estimation of generalized variance when the data are missing from subsets of components are considered. The likelihood ratio test statistic for testing the covariance matrix is equal to a specified matrix, and its asymptotic null distribution is derived when the data matrix is of a monotone pattern. The validity of the asymptotic null distribution and power analysis are performed using simulation. The problem of testing the normal mean vector and a covariance matrix equal to a given vector and matrix is also addressed. Further, an approximate confidence interval for the generalized variance is given. Numerical studies show that the proposed interval estimation procedure is satisfactory even for small samples. The results are illustrated using simulated data.