Premium
Iterative Missing Value Estimation
Author(s) -
Hunt D. N.,
Triggs C. M.
Publication year - 1989
Publication title -
journal of the royal statistical society: series c (applied statistics)
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.205
H-Index - 72
eISSN - 1467-9876
pISSN - 0035-9254
DOI - 10.2307/2348059
Subject(s) - value (mathematics) , estimation , statistics , mathematics , missing data , econometrics , computer science , economics , management
SUMMARY This paper examines iterative methods for estimating missing values in a general designed experiment having a single error term in the analysis of variance. Both the method of Healy and Westmacott and the improved Healy‐Westmacott method of Pearce and others are identified as special cases of successive overrelaxation techniques used in the numerical solution of linear equations. The improved Healy‐Westmacott method is shown to diverge under certain specified conditions. Optimal relaxation parameters are given which guarantee, and in some cases accelerate, convergence. Rates of convergence are compared for selected Latin square designs with missing data. A known disadvantage of iterative methods is their failure to give warning of the confounding which can arise from degenerative configurations of missing values. An extension of the iteration is suggested which enables such confounding to be detected.