Main Page | Report this Page
 
Loading..
   
Science Forum Index  »  Space - Consult Forum  »  MSET methodology
Page 1 of 1    
Author Message
econstatsguy
Posted: Mon Jan 08, 2007 3:04 am
Guest
One of the hot topics at my work in the SPC area these days is the
Smartsignal's Mutivariate State Estimation Technique or MSET, which
comes from the Argonne Lab via Kenneth Gross. Unfortunately, since the
technique is cloaked with a patent, its operation is unclear. See for
example a discussion on
http://swig.stanford.edu/~fox/cs444a/miraclecure.html . I wonder if
anyone here has an interest in this or opinions to share.
Back to top
Old Mac User
Posted: Mon Jan 08, 2007 10:43 am
Guest
I agree with Mike Jordan's analysis. I emphasize there are no free
lunches".

The only rational way to test these "miracles" is to set up a
simulation system... generate synthetic data from a know model and
known random error... pump those "data" through the "new miracle"... do
this hundreds/thousands of times... and from this learn the rate of
false positives and failures to detect. Sadly, in most instances the
"new miracle" is expensive to acquire or you have to access it via the
web, making tests of it difficult if not impossible.

I'm always doubtful of a glaobal (this does everything you'll ever need
in one nice package) solutions.

Let the buyer beware. OMU



econstatsguy wrote:
Quote:
One of the hot topics at my work in the SPC area these days is the
Smartsignal's Mutivariate State Estimation Technique or MSET, which
comes from the Argonne Lab via Kenneth Gross. Unfortunately, since the
technique is cloaked with a patent, its operation is unclear. See for
example a discussion on
http://swig.stanford.edu/~fox/cs444a/miraclecure.html . I wonder if
anyone here has an interest in this or opinions to share.
Back to top
 
Page 1 of 1       All times are GMT - 5 Hours
The time now is Wed Oct 08, 2008 6:39 am
Science Forum Index

Powered by phpBB & ML | Disclaimer



Contact Us