Ja. Fessler, MEAN AND VARIANCE OF IMPLICITLY DEFINED BIASED-ESTIMATORS (SUCH AS PENALIZED MAXIMUM-LIKELIHOOD) - APPLICATIONS TO TOMOGRAPHY, IEEE transactions on image processing, 5(3), 1996, pp. 493-506
Many estimators in signal processing problems are defined implicitly a
s the maximum of some objective function, Examples of implicitly defin
ed estimators include maximum likelihood, penalized likelihood, maximu
m a posteriori, and nonlinear least squares estimation. For such estim
ators, exact analytical expressions for the mean and variance are usua
lly unavailable, Therefore, investigators usually resort to numerical
simulations to examine properties of the mean and variance of such est
imators, This paper describes approximate expressions for the mean and
variance of implicitly defined estimators of unconstrained continuous
parameters, We derive the approximations using the implicit function
theorem, the Taylor expansion, and the chain rule, The expressions are
defined solely in terms of the partial derivatives of whatever object
ive function one uses for estimation, As illustrations, we demonstrate
that the approximations work well in two tomographic imaging applicat
ions with Poisson statistics, We also describe a ''plug-in'' approxima
tion that provides a remarkably accurate estimate of variability even
from a single noisy Poisson sinogram measurement, The approximations s
hould be useful in a wide range of estimation problems.