198 
Fishery Bulletin 11 6(2) 
Assessment year 
Figure 4 
The root mean square error (RMSE) for spawning biomass (SB) by assess¬ 
ment year for each case, (A) time invariant and (B) time varying, and for 
each data scenario (full data, reduced data, eliminated data). The scale of 
the y-axis is the same for comparability of results between the time-invari¬ 
ant and the time-varying simulations. The time-varying eliminated data sce¬ 
nario peaked in year 68 at 221% RMSE. 
the end of the management period (Fig. 3G; note that 
the term median unbiased is used to define cases in 
which the median of the relative errors equals zero). In 
contrast, the median of the estimates of steepness for 
the reduced data scenario were greater than the true 
steepness during the management period (Fig. 3H). 
The eliminated data scenario had the highest among- 
simulation variability among estimates of steepness 
during the management period (Fig. 31) as a result of 
the mixture of simulations in which stocks had rebuilt 
and not rebuilt. 
Reduction or elimination of data during rebuilding 
increased the among-simulation variability in esti¬ 
mates of the size at maximum fishery selectivity and 
the median estimates were generally equal to the true 
value for all data scenarios (Fig. 3, J-L). The among- 
simulation variability of the estimates for the reduced 
and eliminated scenarios improved when the majority 
of the simulated stocks were estimated to be rebuilt 
and fishery composition sample sizes returned to his¬ 
torical levels. The full and reduced data scenarios were 
allowed to estimate dome-shaped selectivity during the 
rebuilding period and resulted in median estimates of 
the width at maximum selectivity that exceeded the 
true values and were highly variable among simula¬ 
tions at the start of the management period (Fig. 3, M 
and N). The eliminated data scenario did not allow es¬ 
timation of dome-shaped selectivity because of the ab¬ 
sence of fishery composition data. The estimates from 
the full and reduced data scenarios for the width at 
maximum selectivity that exceeded the true values for 
this parameter indicated that the data available were 
not sufficient to inform the estimation method about 
the severity of the dome shape in the selectivity curve 
during rebuilding. A higher estimated value indicates 
that the dome in selectivity occurs at larger sizes with 
a higher proportion of the population in relation to the 
operating model at full selectivity. The full data sce¬ 
nario resulted in markedly improved estimates of the 
shape of the dome over the management period, com¬ 
pared with estimates in the reduced data scenario (Fig. 
3, M and N). 
The RMSE for the estimated spawning biomass 
for each assessment year shows the increased preci¬ 
sion of the full data scenario during the rebuilding 
period compared with that of the reduced and elimi¬ 
nated data scenarios (Fig. 4A ). The eliminated data 
scenario resulted in the highest RMSE over the entire 
management period (Fig. 4A). However, the RMSE for 
the reduced data scenario improved over the manage¬ 
ment period as simulated stocks began to be assessed 
as rebuilt to the target biomass and as sample sizes re¬ 
turned to historical levels. The limited improvement in 
the RMSE for the eliminated data scenario was driven 
by the simulations in which the stocks never were pro¬ 
jected to rebuild to the target biomass (35 out of 100 
simulations). 
In the absence of data collection, the performance of 
the estimation method was dependent upon the ability 
of the historical data to inform parameter estimates. 
An examination of the eliminated data scenario more 
closely revealed a pattern in the performance of the 
estimation method based on the estimation of steep¬ 
ness in the first assessment year. The eliminated data 
scenario simulations were divided and plotted on the 
basis of whether the estimation method projected the 
stock in the simulation to rebuild (65 simulations) or 
to fail to rebuild (35 simulations) by the end of the 
management period. To allow comparison between the 
eliminated and the full data scenarios, the estimates 
from the full data scenario were also divided into the 
same 2 groups and plotted. The estimates of spawning 
