One replicate's recruitment deviations, the fit's first and drawn from its penalty after

Deviations over the first n_cond_yrs years are the fit's. After them, every deviation the fit penalizes is drawn from the penalty get_rec_devs_penalty puts on it: independent deviations about minus half their variance times the year's bias ramp, or about the fit's own mean under RecDevs_pen_center = 1; a random walk from the previous calendar year; or an AR1 about its centered mean. Each sd is divided by the square root of the cell's Wt_Rec, since a weighted penalty is the same density at that smaller sd.

Usage

rec_devs_past_fit(data, pars, rep, n_cond_yrs, n_yrs, map_RecDevs = NULL)

Arguments

data, pars, rep

The fit's data list, parameter list and report, or one replicate's view of them.

n_cond_yrs

Years whose deviations stay at the fit's.

n_yrs

Years the operating model runs.

map_RecDevs

The fit's map for ln_RecDevs, whose NA cells are fixed values rather than a process, or NULL when every cell is estimated.

Value

Array [n_pop, n_regions, n_yrs] of log recruitment deviations.

Details

A deviation the fit leaves unpenalized keeps the fit's value: one mapped off, the first years under dont_pen_recdev_first, a walk's first year under its diffuse start, a zero weight, or a region with no recruits. Cells sharing a penalty level take one draw, and years past the fit's deviations take none, as in the estimation model.