<div dir="ltr">Mark,<div><br></div><div>The approach I take for stable estimation of F rates is to use a well-defined starting point for the search and to search over a fixed number of iterations. In particular, I get the starting point from Pope's approximation to calculate an exploitation rate (for each fleet relative to the mid-year available biomass for that fleet); then convert these U's into an approximation for each F, then tune these F's over a fixed number of iterations. This converges well enough with just 4 iterations even with multiple fleets. The trick to rapid convergence is to base the updated F values based on the anticipated total Z after adjusting the Fs, not the current total Z. Note that this is in a separable model, so it is looking for the F that matches catch biomass conditioned on a set of selectivity parameters.<div>
<br></div><div style>This approach leads to very steep gradients for other parameters when F is high and the catch data is considered to be precise, so convergence of the overall model is slow because the model is constantly maintaining a good fit to the catch data. For these cases, I find it useful to switch over (during BETWEEN_PHASES) to treating the F's as ADMB model parameters, rather than as a set of scaling factors to match the catch. When treating the F's as parameters, the model tends to wait until the last iterations to get a good fit to the catches.</div>
<div style><br></div><div style>Overall, I find that the variance on other model outputs (such as final population biomass) is rather insensitive to F as coefficient vs. F as parameter, but haven't tested this in high F situations.</div>
<div style><br></div><div style>Rick</div></div></div><div class="gmail_extra"><br><br><div class="gmail_quote">On Mon, Mar 11, 2013 at 10:35 AM, Fowler, Mark <span dir="ltr"><<a href="mailto:Mark.Fowler@dfo-mpo.gc.ca" target="_blank">Mark.Fowler@dfo-mpo.gc.ca</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Yes, and I use epsilon for proper parameters. The situation I'm<br>
addressing is 'non-ADMB' iterative mortality (F) estimates. These are<br>
just loop functions to estimate F's to create population and mortality<br>
matrices. Obviously candidates as parameters, but I needed to keep them<br>
as originally written. I was translating a particular type of VPA<br>
written in a language called ACON to R and ADMB. Only the reference year<br>
abundances are estimated with a serious optimizing function in the<br>
original model (to reasonably estimate catchabilities). Initially this<br>
served for truthing until I finished the translations. Hence the issue<br>
of dealing with extremely small values are the profusion of F loops. And<br>
I'm doing these in an equivalent manner to your suggestion, I just don't<br>
use epsilon per se because I like to restrict it to parameters.<br>
<br>
I've since been keeping the loops while comparing vulnerabilities to<br>
local minima across the three programs, so only the reference year<br>
abundances are optimized directly. Both optim and the ACON optimizer<br>
(NLLS) clearly work left to right on parameter sets. The optim function<br>
gets trapped in local minima with high uninformative constants for<br>
priors on a vector that grades high to low. Starting left-most they go<br>
up, which is correct relative to numbers to the right but wrong in<br>
absolute terms. NLLS starts out similarly (run with few iterations you<br>
get the same answer) but doesn't stay there, so maybe breaks the<br>
left-to-right rule at some point to explore the range, or has some<br>
threshold point at which it gives up on the first parameter. So far I'm<br>
stymied with ADMB. I'm thinking maybe the F estimations confound<br>
tracking of gradients?<br>
<br>
<br>
> Mark Fowler<br>
Population Ecology Division<br>
> Bedford Inst of Oceanography<br>
> Dept Fisheries & Oceans<br>
> Dartmouth NS Canada<br>
B2Y 4A2<br>
Tel. <a href="tel:%28902%29%20426-3529" value="+19024263529">(902) 426-3529</a><br>
Fax <a href="tel:%28902%29%20426-9710" value="+19024269710">(902) 426-9710</a><br>
Email <a href="mailto:Mark.Fowler@dfo-mpo.gc.ca">Mark.Fowler@dfo-mpo.gc.ca</a><br>
Home Tel. <a href="tel:%28902%29%20461-0708" value="+19024610708">(902) 461-0708</a><br>
Home Email <a href="mailto:mark.fowler@ns.sympatico.ca">mark.fowler@ns.sympatico.ca</a><br>
<div class="HOEnZb"><div class="h5"><br>
<br>
-----Original Message-----<br>
From: <a href="mailto:users-bounces@admb-project.org">users-bounces@admb-project.org</a><br>
[mailto:<a href="mailto:users-bounces@admb-project.org">users-bounces@admb-project.org</a>] On Behalf Of dave fournier<br>
Sent: March 8, 2013 11:52 AM<br>
To: <a href="mailto:users@admb-project.org">users@admb-project.org</a><br>
Subject: Re: [ADMB Users] very small number math<br>
<br>
This is proabably a good use of phases. In the early phase you add<br>
something<br>
<br>
like<br>
<br>
x/(eps+y)<br>
<br>
and let eps get smaller or 0 in a later phase.<br>
_______________________________________________<br>
Users mailing list<br>
<a href="mailto:Users@admb-project.org">Users@admb-project.org</a><br>
<a href="http://lists.admb-project.org/mailman/listinfo/users" target="_blank">http://lists.admb-project.org/mailman/listinfo/users</a><br>
_______________________________________________<br>
Users mailing list<br>
<a href="mailto:Users@admb-project.org">Users@admb-project.org</a><br>
<a href="http://lists.admb-project.org/mailman/listinfo/users" target="_blank">http://lists.admb-project.org/mailman/listinfo/users</a><br>
</div></div></blockquote></div><br><br clear="all"><div><br></div>-- <br><b style="font-weight:normal"><span style="font-family:Arial;font-size:13px;color:rgb(102,102,102);font-weight:bold;vertical-align:baseline;white-space:pre-wrap">Richard D. Methot Jr. Ph.D.</span><br>
</b><b style="font-weight:normal"><span style="font-family:Arial;font-size:13px;color:rgb(102,102,102);font-style:italic;vertical-align:baseline;white-space:pre-wrap">NOAA Fisheries - </span></b><b style="font-weight:normal"><font color="#666666" face="Arial"><span style="white-space:pre-wrap"><i>Science Advisor for Stock Assessments</i></span></font></b><div>
<b style="font-weight:normal"><span style="font-family:Arial;font-size:13px;color:rgb(102,102,102);vertical-align:baseline;white-space:pre-wrap">Office: 206-860-3365</span></b><b style="font-weight:normal"><br><span style="font-family:Arial;font-size:13px;color:rgb(102,102,102);vertical-align:baseline;white-space:pre-wrap">Mobile: 301-787-0241</span><br>
</b></div>
</div>