Imputation for Skewed Data: Multivariate Lomax Case

Most multiple imputation methods for multivariate missing data have been developed for normally distributed data. However, methods may not be suitable for nonnegative and/or highly skewed data. We propose an approach by using Expectation-Maximization (EM) method based on the assumption of multivaria...

Full description

Saved in:
Bibliographic Details
Published in:Sankhyā. Series B (2008) 2021-05, Vol.83 (Suppl 1), p.86-113
Main Authors: Lun, Zhixin, Khattree, Ravindra
Format: Article
Language:English
Subjects:
Mathematics and Statistics
Statistics
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Most multiple imputation methods for multivariate missing data have been developed for normally distributed data. However, methods may not be suitable for nonnegative and/or highly skewed data. We propose an approach by using Expectation-Maximization (EM) method based on the assumption of multivariate Lomax distribution on non-negative skewed data. Extensive simulations show that this proposed method outperforms the regular normality-based EM and k -nearest-neighbor ( k NN) imputation methods under the missing completely at random (MCAR) mechanism. An application on a real-world biomedical data is then provided.
ISSN:0976-8386
0976-8394
DOI:10.1007/s13571-021-00251-4