troubleshooting catch/F scaling issues

[Brooke Lowman]

Hello,

I am working on getting an initial functional run with subset of my available datasets for an Atlantic croaker assessment: one commercial catch series, one recreational removals series (discards included as catch), and two YOY surveys. I am struggling to get anywhere close to fitting to the catch, and I'm unsure of why. The commercial catch estimates (and F) are forced near zero in almost all years. The recreational removals series is also poorly estimated. However, the magnitude of recreational removals is about right, and the recreational F estimate is near the max of 4 in many years.

I thought that my initial equilibrium assumption could be causing an issue, but I tried several adjustments (fixed at 0.1, allowing estimation, increasing CV), but none yielded any improvement. Adjusting retention parameter for the commercial fleet did not appear to help either. 

Any insight into what could be causing the issue would be greatly appreciated. Inputs are attached.

Thank you,

Brooke

[Jason Cope]

Hi Brooke,

One question upon a very quick look at your data file (and not having run the model)- I notice your CV on catch is high (0.1 and 0.3 respectively).

Have you tried a CV of 0.01 or smaller on the catches? This should tell the model to fit the catches really well, whereas what you currently have gives the model a fair bit of leeway not to fit the catches.

-Jason

[Mark Maunder]

What do people think about using a double cumulative normal for an asymptotic selectivity. This would allow a bit more flexibility than using the logistic or a half normal using the double normal.

 

It could simply be added under the double normal definition when the plateau is -999, for example.

 

See equation 2.8 in Julio, J.M., 2007. The Fan Chart …. Borradores de Economia No. 468

 

Let me know if you think this is a good idea.

 

S(i) = C*sqrt(2pi)*sd1*cumN(i,mean,sd1) if i<=mean

S(i) = 1- C*sqrt(2pi)*sd2*(1-cumN(I,mean,sd2)) if i>mean

 

C=sqrt(2/pi)/(sd1+sd2)

 

ADMB has a cumulative normal distribution

 

Let me know what you think,

 

Mark

--
You received this message because you are subscribed to the Google Groups "SS3 - Forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ss3-forum+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ss3-forum/70d49261-d395-4e41-aed1-cd1c3af2aaf9n%40googlegroups.com.

[Brooke Lowman]

Hi Jason,

Thanks for the quick response! I hadn't thought of 0.1 as a high CV (the catches, especially recreational, are pretty uncertain). But you're absolutely right. Setting them all to 0.01 works. Thank you for the suggestion.

Best,
Brooke