Model Identification for Estimating Missing Values in Space-Time Data Series: Monthly Inflation in the U. S. Urban System, 1977–1985

Abstract

Dealing with the problem of missing values in data sets has received increasing attention by applied statisticians over the last decade (see Bennett, et. al., 1984). While there are many conceptual and practical statistical reasons for developing ways of coping with this particular problem, the motivation for work reported in this paper stems from a need to interpolate missing data values in order to calibrate optimization models, and other sorts of non-statistical models, as was done in Griffith (1983). In contrast, for the most part statisticians have been concerned with obtaining unbiased parameter estimates in the presence of missing data. Hence the thrust of this paper is more closely akin to the aims and goals of forecasting. Not surprisingly, then, the forecasting literature will be drawn upon heavily, as the primary basis for developments reported in this paper.