Bayesian Treatment of Multivariate Normal Data With Observations Missing at Random

• NCJ Number: 95227
• Author(s): J B Kadane; R L Trader
• Date Published: 1983
• Length: 18 pages
• Grant Number(s): 81-IJ-CX-0087
• Contract Number(s): N00014-82-K-0622
• Sponsoring Agency: US Dept of the Navy; University of Maryland; National Institute of Justice (NIJ); US Dept of Justice NIJ Pub
• Corporate Author: Carnegie Mellon University
• Sale Source: National Institute of Justice/; NCJRS Photocopy Services
• Publication Type: Statistics
• Language: English
• Country: United States of America
• Note: Technical Report number 308

Annotation

This paper provides a Bayesian formulation of the multivariate normal missing data problem when observations are missing at random.

Abstract

Conjugate prior distributions are considered both when the covariance matrix is known and when it is unknown. When the covariance matrix is known, the posterior distribution of the mean is multivariate normal. When the covariance matrix is unknown, the conditional distribution of the mean given the covariance matrix is again multivariate normal. The kernal of the distribution of the covariance matrix is given and is seen to be difficult to integrate analytically. The paper concludes with some comments of the bivariate case, which is partially tractable for nested missing data. Thirteen references are listed. (Author abstract modified)