Statistical aspects of W. D. Fisher's method of optimal aggregation

From MaRDI portal





\textit{W. D. Fisher}'s approach [Econometrica 30, 744-769 (1962; Zbl 0122.154)] to approximating a known linear model \(y=Ax\) by an aggregation- disaggregation sequence involves a moment matrix M which is supposed to be known. This paper investigates the statistical problems arising when M is not known. As an example we consider the open static Leontief model with A being the Leontief inverse. Under the assumption that the exogenous variables are multinormal with zero mean, tests and confidence intervals for the loss of information are given. The use of maximum entropy estimation of M for insufficient data is discussed. Statistical procedures to evaluate the goodness of aggregation-disaggregation of a time series of matrices A are also described.











This page was built for publication: Statistical aspects of W. D. Fisher's method of optimal aggregation

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3745145)