s-news
[Top] [All Lists]

[S] non-linear least squares - summary

To: s-news@wubios.wustl.edu
Subject: [S] non-linear least squares - summary
From: Miles Harkness <mharknes@laurel.ocs.mq.edu.au>
Date: Thu, 24 Jun 1999 17:21:15 +1000 (EST)
Sender: owner-s-news@wubios.wustl.edu
Following up on my question to the list on what FALSE CONVERGENCE meant -
an error message I got when trying to use the function ms - here are the
replies that I received.  Hope they are of help.

>From Bill Venebles
CMIS, Cleveland
It usually means you have started from a very bad starting value, hit a
region where the sum of squares function is essentially flat (so the
minimum
is not defined), the algorithm has detected this and given up with this
warning.  It means "using the starting values you gave me I got lost and
can't find anything remotely like a minimum".

>From Bob Byers
Centers for Disease Control & Prevention
NCHSTP/Statistics & Data Mgmt Branch
        False convergence means that the estimates are not changing
very much but the numerical derivatives are not consistent with having
found a minimum. Usually this means your starting values for parameters 
were not near the minimum, although the problem could be with the 
function you are trying to minimize. I recommend playing around with the
start values.

Thanks for your help.
Miles Harkness

*******************************************************
Miles Harkness            mharknes@laurel.ocs.mq.edu.au
Macquarie University      +61 (02) 9850 8564 (office)
Sydney                    +61 (02) 9460 2182 (home)
*******************************************************





-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu.  To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message:  unsubscribe s-news

<Prev in Thread] Current Thread [Next in Thread>