Normal-distribution-based maximum likelihood (NML) is most widely used for missing data analysis although real data seldom follow a normal distribution. When missing values are missing at random (MAR), recent results indicate that NML estimates (NMLEs) are still consistent for nonnormally...