Geometry of the Cramér-Rao bound (Q1801395)

The problem of identifying parameters, that structure the mean vector in a multivariate normal distribution, is considered. The Fisher matrix (this is a Gramian constructed from the sensitivity vectors that are the first order variation in the mean with respect to the parameters) and its inverse are studied. A procedure for partitioning the parameters into disjoint sets is proposed. The inverse of the similarly partitioned Fisher matrix is Cholesky factored to produce a partitioned inverse that brings some geometrical interpretations. For example, it is the angle between a given sensitivity vector and the linear subspace spanned by all other sensitivity vectors that determines the variance bound for identifying a given parameter. The obtained geometrical interpretations may also be given filtering interpretations in a related multivariate normal estimation problem.