I shared the latest sdmTMB output you had showed me with Eric. He thought it looked good and thought the spatial effects were particularly cool! Here are a couple of suggestions he had:
Try fitting year as a fixed effect (in addition to the two ways you are already doing it). He said (journal) reviewers would likely ask for that. So, you could compare the AIC values for models with year as a fixed effect, year as a random effect, and year as a random walk and then choose based on which gave the best fit.
Investigate reducing the number of knots in the weight at age GAM so it is less "wiggly", essentially forcing it to asymptote.
Try fitting a 2d smooth for cohort and age to allow for interactions between them. So far, I think we've assumed that the cohort effect would just move the weight at age curve up and down, but not change its shape. But it might make more sense biologically to think that the shape might change with different cohorts (weight at age increases more quickly with age but asymptotes to similar values). I think you did some explorations with 2d gams initially before you switched to sdmTMB, but I don't remember the details.
Fitting more models might not be your top priority at the moment but wanted to post these here so we have them to return to later. Eric's also happy to meet with you/us to chat more down the road as desired.
I shared the latest sdmTMB output you had showed me with Eric. He thought it looked good and thought the spatial effects were particularly cool! Here are a couple of suggestions he had:
Try fitting year as a fixed effect (in addition to the two ways you are already doing it). He said (journal) reviewers would likely ask for that. So, you could compare the AIC values for models with year as a fixed effect, year as a random effect, and year as a random walk and then choose based on which gave the best fit.
Investigate reducing the number of knots in the weight at age GAM so it is less "wiggly", essentially forcing it to asymptote.
Try fitting a 2d smooth for cohort and age to allow for interactions between them. So far, I think we've assumed that the cohort effect would just move the weight at age curve up and down, but not change its shape. But it might make more sense biologically to think that the shape might change with different cohorts (weight at age increases more quickly with age but asymptotes to similar values). I think you did some explorations with 2d gams initially before you switched to sdmTMB, but I don't remember the details.
Fitting more models might not be your top priority at the moment but wanted to post these here so we have them to return to later. Eric's also happy to meet with you/us to chat more down the road as desired.