LISTSERV mailing list manager LISTSERV 15.5

Help for FISH-SCI Archives

FISH-SCI Archives

FISH-SCI Archives


Next Message | Previous Message
Next in Topic | Previous in Topic
Next by Same Author | Previous by Same Author
Chronologically | Most Recent First
Proportional Font | Monospaced Font


Join or Leave FISH-SCI
Reply | Post New Message
Search Archives

Subject: Re: Likelihood ratio tests for VBGFs
From: Steve Branstetter <[log in to unmask]>
Reply-To:Academic forum on fisheries ecology and related topics <[log in to unmask]>
Date:Tue, 2 Jul 1996 13:13:20 GMT

text/plain (44 lines)

On Jul 02, 1996 14:32:01, '"L. Vilizzi"
<[log in to unmask]>' wrote:
>I wonder if anybody can give me some advice on likelihood ratio tests for
>statistical comparison of growth curves.
>I would greatly appreciate any feedback on this topic,
Obviously, fitting the vB curve is relatively simple; numerous packages
exist - FISHPARM used to be quite a standard, or any of the current stat
packages such as systat, sigmastat, statgraph, or just plain old nonlin SAS
will produce a curve and parameter estimates.
Comparing the parameters to test for differences comes in a variety of
methods too.  SAS will do comparisons of curves, and when I was doing A/G
work, there was a method I used described by Bernard (1981?).  Since then,
as you note there have been several papers of techniques.   A recent A/G
paper by Sminkey and Musick 1995 Copeia (#4, I think) {the results of
Sminkey's dissertation) used several different methods to compare between
curves generated for two different time periods.  You might review that
paper as well.
The real problem with the vB curve is the relationship among the variables;
as Lmax goes up, K goes down, etc.; they aren't independent.  In my own
case (Branstetter et al 1987, Fish. Bull. 85:265-275), Bernard's test
showed significant differences between two curves, yet the data points
overlapped between the two datasets.  Significance seems to come readily
with these parameters and is probably driven more by the data set itself
(whether its biased by a specific size/age group, lack of big or small
fish, or just limited in number period).  IMHO, you will get statistical
signficance even if its not a biologically significant difference, thus
what have you really proved?  That the two populations really are
different, or that you can find and use a computer program that says they
are different?  Just don't forget the original purpose was to describe a
life history trait for a species.
Steve Branstetter, Ph.D., Program Director
Gulf & South Atl. Fisheries Develop. Fndn.
Suite 997 Lincoln Center, 5401 W. Kennedy
Tampa, FL  33609
ph:813-286-8390      fax: 813-286-8261
e-mail:  [log in to unmask]

Back to: Top of Message | Previous Page | Main FISH-SCI Page



CataList Email List Search Powered by the LISTSERV Email List Manager