modeling parameters in stock synthesis modeling population processes 2009 iattc workshop
TRANSCRIPT
Modeling Parameters in Stock Synthesis
Modeling population processes2009 IATTC workshop
Outline
• General framework
• Bounds and priors
• Temporal variation
• Relationship among parameters
General framework
• All parameter inputs have 14 or 7 elements
• First 7: bounds, init value, prior info, phase
• Next 7: advanced options for time variation
• Conditional inputs depending on options #_LO HI INIT PRIOR PR_type SD PHASE env-var use_dev dev_minyr dev_maxyr dev_stddev Block Block_Fxn # Label 0.05 0.15 0.1 0.1 0 0.8 -3 0 0 0 0 0.5 0 0 # NatM_p_1_Fem_GP_1 -3 3 0 0 0 0.8 -3 0 0 0 0 0.5 0 0 # NatM_p_2_Fem_GP_1 10 45 36.0 36.0 0 10 2 0 0 0 0 0.5 0 0 # L_at_Amin_Fem_GP_1 40 90 70.0 70.0 0 10 2 0 0 0 0 0.5 0 0 # L_at_Amax_Fem_GP_1 0.05 0.25 0.15 0.15 0 0.8 3 0 0 0 0 0.5 0 0 # VonBert_K_Fem_GP_1 0.05 0.25 0.1 0.1 0 0.8 -3 0 0 0 0 0.5 0 0 # CV_young_Fem_GP_1 -3 3 0.25 0.25 0 0.8 -3 0 0 0 0 0.5 0 0 # CV_old_Fem_GP_1 -3 3 0 0 0 0.8 -3 0 0 0 0 0.5 0 0 # NatM_p_1_Mal_GP_1 -3 3 0 0 0 0.8 -3 0 0 0 0 0.5 0 0 # NatM_p_2_Mal_GP_1 -3 3 0 0 0 0.8 -3 0 0 0 0 0.5 0 0 # L_at_Amin_Mal_GP_1 -3 3 0 0 0 0.8 -2 0 0 0 0 0.5 0 0 # L_at_Amax_Mal_GP_1 -3 3 0 0 0 0.8 -3 0 0 0 0 0.5 0 0 # VonBert_K_Mal_GP_1 -3 3 0 0 0 0.8 -3 0 0 0 0 0.5 0 0 # CV_young_Mal_GP_1 -3 3 0.25 0.25 0 0.8 -3 0 0 0 0 0.5 0 0 # CV_old_Mal_GP_1 -3 3 2.0e-06 2.0e-06 0 0.8 -3 0 0 0 0 0.5 0 0 # Wtlen_1_Fem -3 4 3.0 3.0 0 0.8 -3 0 0 0 0 0.5 0 0 # Wtlen_2_Fem 50 60 55 55 0 0.8 -3 0 0 0 0 0.5 0 0 # Mat50%_Fem -3 3 -0.25 -0.25 0 0.8 -3 0 0 0 0 0.5 0 0 # Mat_slope_Fem -3 3 1 1 0 0.8 -3 0 0 0 0 0.5 0 0 # Eg/gm_inter_Fem
Bounds and priors
• All parameters bounded
• Prior options: uniform, normal, lognormal, symmetric and non-symmetric beta
Parameter value
Pri
or
de
nsi
ty
Pmin Pmax
UniformNormalLog-normalSymmetric betaNon-symmetric beta
Soft bounds
• Optional penalty (set in starter file) applied to all parameters
• Keeps ADMB from getting stuck on bounds
• Acts along with user-specified priors
• Equivalent to symmetric beta with shape parameter = 0.001
Parameter value
-lo
g(L
)
0.0 0.2 0.4 0.6 0.8 1.0
0.0
00
0.0
05
0.0
10
Temporal variationDeviations (N std. dev. pars.)
Random walk (N -1 std. dev. pars.)
Blocks (1 par. per block)
Trend (3 pars.)
Temporal variation: blocks
• Requires conditional input for extra parameters lines (same as other variation types)
• Fixed time intervals specified in control file• Additional parameters may be:
– Multiplicative offset from base value– Additive offset from base value– Replace base value for interval of years– May have random walk from one block to next
Temporal variation: deviations
Temporal variation: random walk
• Defined by – Type (base+dev or base∙edev) – Start and end years for – Normal distribution penalty
• Not zero-centered
• Similar to deviations, but one fewer parameter
• Parameters represent differences• Normal distribution penalty
Temporal variation: trends
• Only 3 parameters
• Smooth alternative to blocks for cases that don’t support many parameters
• Final value may be offset from base or new value
Parameter as function of covariate
• Environmental variable: Ey– Pary = base+link∙Ey or base∙eEy
– May be combined with other options (i.e. deviations around environmental index)
• Covariate relationship to be used in future versions of SS for density dependence:– Mortality parameters as a function of biomass
Keeping time-varying parameters within bounds
Options:
• time varying parameters unconstrained by bounds on base parameter
• logistic transformation to keep adjusted parameter value within bounds of base
Offsets from other parameters
• Parameters for males often treated as offsets from females– growth– mortality– selectivity
• Additive or multiplicative options• Makes hypothesis testing easy• To be covered in more detail in upcoming
sessions of IATTC workshop