The theorems and other results in the following paper
are correct, as far as I know, but
there is a step missing in the proof of Theorem 1.
Kass, R.E., Tierney, L. and Kadane, J.B. (1990) The validity of posterior expansions based on
Laplace's method, Essays in Honor of George Bernard,
eds. S. Geisser, J.S. Hodges, S.J. Press, and A. Zellner, Amsterdam:
North Holland, 473-488.
The validity of posterior expansions based on Laplace's method
must consider (1) local regions vanishingly close to the mode (where
expansions take place), (2) distant regions (e.g., the exterior of a
ball about the mode having fixed radius, which will have
exponentially-decreasing probability), and (3) non-distant regions
(e.g., the interior of the ball in (2)). In the Kass, Tierney, and
Kadane paper
we neglected to spell out what happens in (3). The argument
is not hard, and depends on the assumed positive definiteness of the
Hessian matrix. For those who would like details,
the oversight was corrected in Theorem 2.2.13 and Lemma 2.2.16 of my
book, Kass and Vos, "The Geometry of Asymptotic Inference"
(Wiley). The context there is curved exponential families but the
method is the same.
|