plotRealiszation.Rmd shows that the Larkin and Ricker outperform EDM in terms of absolute forecast error in recruitment, for when beta parameters are even (0.25, 0.25, 0.25, 0.25) and when beta0 is dominant (0.45, 0.25, 0.15, 0.15).
I suggest we consider
Forecasting returns by return year (your original suggestion). I have coded this in for the Larkin model, but have assumed that the average age proportions are known (p_prime)
Adding an option for time-varying alpha parameters in the underling simulation (e.g., cyclic or persistent trend). There is evidence for time-varying productivity (alpha value) among Fraser Sockeye, and plausible biological mechanisms related to oceanography and climate impacts. This may be more realistic, and may cause deterioration of Larkin performance.
plotRealiszation.Rmd shows that the Larkin and Ricker outperform EDM in terms of absolute forecast error in recruitment, for when beta parameters are even (0.25, 0.25, 0.25, 0.25) and when beta0 is dominant (0.45, 0.25, 0.15, 0.15). I suggest we consider