The TMB parameter vector uses transformed, abbreviated names,
log_M1, R_log_sd, index_log_q. This returns the table mapping each one
to the quantity it represents on its natural scale, which process it belongs
to, what it means, and its dimensions.
Arguments
- internal
Character vector of internal parameter names, as they appear in
names(fitted$estimated_params)or in ansdreport(). DefaultNULLreturns every parameter.- process
Character vector restricting the table to one or more parts of the model:
"recruitment","mortality","growth","fishing","catchability","selectivity","observation","predation","linkage"or"internal".
Value
A data frame with one row per parameter and columns internal,
natural, process, meaning and dims.
Details
A name beginning log_, or ending _log_sd, is estimated on the log scale,
so exp() turns it into the quantity named in the natural column. Most
fitted values are already back-transformed in fitted$quantities, so this is
mainly needed when reading estimated_params or an sdreport() directly.
Dimensions use the model's own notation: nspp species, nsex sexes,
nages age bins, nyrs years including the projection, nyrs_hind hindcast
years only, n_flt fleets and n_sel selectivity blocks. n_sel equals
n_flt: fleets sharing a Selectivity_index share one block through the
map, not through a narrower array.
See also
fit_mod() for the fitted object these names come from.
Examples
# The whole table
head(parameter_dictionary())
#> internal natural process
#> 1 dummy dummy internal
#> 2 log_pop_scalar pop_scalar internal
#> 3 rec_pars R0 / alpha / beta recruitment
#> 4 rec_dev recruitment deviations recruitment
#> 5 R_log_sd sigma_R recruitment
#> 6 init_dev initial-age deviations recruitment
#> meaning
#> 1 Placeholder parameter; the only free parameter under estimateMode = 4.
#> 2 Multiplier on user-supplied numbers-at-age; estimated for estDynamics = 2 under predation, 1 otherwise.
#> 3 Stock-recruit parameters: mean recruitment (or R0), and the SRR alpha and beta.
#> 4 Annual log-scale recruitment deviations.
#> 5 Standard deviation of the recruitment deviations; estimated only when random_rec = TRUE.
#> 6 Deviations defining the initial (first-year) age structure; one per age from minage + 1 upward, since age minage in the first year is recruitment.
#> dims
#> 1 [1]
#> 2 [nspp]
#> 3 [nspp, 3]
#> 4 [nspp, nyrs]
#> 5 [nspp]
#> 6 [nspp, nages-1]
# What is R_log_sd?
parameter_dictionary("R_log_sd")
#> internal natural process
#> 1 R_log_sd sigma_R recruitment
#> meaning
#> 1 Standard deviation of the recruitment deviations; estimated only when random_rec = TRUE.
#> dims
#> 1 [nspp]
# Everything governing selectivity
parameter_dictionary(process = "selectivity")
#> internal natural process
#> 1 log_sel_slp selectivity slope selectivity
#> 2 sel_inf selectivity inflection selectivity
#> 3 log_sel_apical selectivity apical height selectivity
#> 4 sel_dn6 SS3 double-normal parameters selectivity
#> 5 log_sel_slp_dev slope deviations selectivity
#> 6 sel_inf_dev inflection deviations selectivity
#> 7 sel_coff selectivity coefficients selectivity
#> 8 sel_coff_dev coefficient deviations selectivity
#> 9 sel_dev_log_sd sigma_sel selectivity
#> 10 sel_curve_pen selectivity penalty weights / AR1 correlations selectivity
#> meaning
#> 1 Logistic-family selectivity slope; row 1 ascending, row 2 descending.
#> 2 Logistic-family age/length at 50% selection; row 1 ascending, row 2 descending.
#> 3 Log multiplier on one sex's whole curve, applied after the form and before normalization; 0 = no offset. Estimated only through a selectivity linkage on `apical`.
#> 4 DoubleNormalSS3 (Stock Synthesis size pattern 24) parameters on SS3's own scales: peak (cm or age), logit top width, log ascending width, log descending width, logit initial and logit final selectivity. A starting value of -999 on either of the last two is SS3's switch for leaving that end unscaled: the parameter then drops out of the curve and is fixed. Time variation is through selectivity linkages.
#> 5 Annual deviations on the selectivity slope.
#> 6 Annual deviations on the selectivity inflection point.
#> 7 Non-parametric selectivity-at-bin coefficients.
#> 8 Annual deviations on the non-parametric selectivity coefficients.
#> 9 Standard deviation of the time-varying selectivity deviations.
#> 10 Meaning depends on the fleet's Selectivity. For the non-parametric forms these are weights on the shape and curvature penalties, supplied via fleet_control and not estimated. For '2DAR1' and '3DAR1' the same slots hold estimated correlations: slot 1 across selectivity BINS, slot 2 across YEARS, and for 3DAR1 slot 3 across COHORTS.
#> dims
#> 1 [2, n_sel, nsex]
#> 2 [2, n_sel, nsex]
#> 3 [n_sel, nsex]
#> 4 [6, n_sel, nsex]
#> 5 [2, n_sel, nsex, nyrs_hind]
#> 6 [2, n_sel, nsex, nyrs_hind]
#> 7 [n_sel, nsex, n_sel_bins]
#> 8 [n_sel, nsex, n_sel_bins, nyrs_hind]
#> 9 [n_sel]
#> 10 [n_sel, 3]
# Search the meanings by keyword
dict <- parameter_dictionary()
dict[grep("catchability", dict$meaning, ignore.case = TRUE), ]
#> internal natural process
#> 27 index_log_q q catchability
#> 29 index_q_dev q deviations catchability
#> 30 index_q_log_sd sigma_q_prior catchability
#> 31 index_q_dev_log_sd sigma_q_dev catchability
#> meaning
#> 27 Survey/index catchability.
#> 29 Annual deviations on catchability when q is time-varying.
#> 30 Standard deviation of the prior on catchability (used when Estimate_q = 2).
#> 31 Standard deviation of the time-varying catchability deviations.
#> dims
#> 27 [n_flt]
#> 29 [n_flt, nyrs_hind]
#> 30 [n_flt]
#> 31 [n_flt]