264 Fishery Bulletin 119(4) 
A 2-way analysis of variance (ANOVA) was 
used to assess average MIRs between different 
J: R times of year and among age bins, as recom- 
GAT... mended by Campana (2001). The Akaike infor- 
OL mation criterion (AIC) was used to identify 
the best model for analysis, varying predictors 
Gulf (Month Bin, Age Bin, and Region), additivity, 
of Maine and interactions. Month bins (e.g., January— 
February and March—April) were used instead 
of individual months because missing data 
would have precluded interactive models from 
running. Missing data were also the reason that 
region (i.e., capture location) could be included 
only as an additive predictor and not as an 
interactive one. The regions designated for this 
analysis were north and south of the Hudson 
Canyon, as described in the report for the latest 
stock assessment (NEFSC, 2017). This canyon 
begins 100 km from the mouth of the Hudson 
River and extends approximately 600 km to 
the southeast (NEFSC, 2017). Additionally, 
monthly average MIRs were used to visually 
assess the timing of annulus formation for each 
age bin. 
The possibility of differences in growth of 
black sea bass between regions (Dery and Mayo, 
1988), as well as the recent separation of the 
northern stock into 2 subunits north and south 
of the Hudson Canyon, motivated an analysis 
that included Region as an interactive predictor. 
In a 3-way ANOVA, seasons were used instead 
of month bins because of missing data. Sea- 
sons, chosen on the basis of information avail- 
able about migration of black sea bass (they 
arrive inshore by April and leave by October 
or November; Drohan et al., 2007), were as fol- 
lows: January, February, and March (winter); 
April, May, and June (spring); July, August, and 
Sample size 
Atlantic Ocean 
NO In) _—_—__n_] 
74°W 72°W 
Figure 1 
Map of locations where black sea bass (Centropristis striata) were 
caught in the Atlantic Ocean off the northeastern United States 
during 2013-2017. Sagittal otoliths were removed from sampled fish 
and used for marginal increment analysis and first annulus valida- 
tion. Size of circles indicates the number of samples collected at each 
location. 
et al., 2017), as the MIR (Vilizzi and Walker, 1999; Zlokovitz 
et al., 2003). If annuli are formed once per year, monthly 
MIR should indicate a sinusoidal pattern with only one 
minimum per year, when annulus formation is complete 
and new growth begins (Wenner et al., 1986; Pilling et al., 
2000). Marginal increment ratios were calculated following 
Condini et al. (2014) by dividing the marginal increment 
(completed edge growth) by the measurement of the pre- 
sumed previous year’s growth (full band pair, one translu- 
cent band and one opaque band) (Fig. 3B): 
where RF, = the otolith radius (core to edge); 
R,_, = the measurement from otolith core to the dis- 
tal edge of the last opaque band; and 
R,_2 = the measurement from otolith core to the dis- 
tal edge of the penultimate opaque band. 
September (Summer); and October, November, 
and December (fall). Regions were designated 
as described previously, although a difference 
in sample sizes should be noted (north: n=970; south: 
n=365). Methods used to account for this unbalanced 
design are described at the end of this section. The AIC 
was used to identify the best model for analysis, varying 
predictors (Season, Age Bin, and Region), additivity, and 
interactions. 
Because an ANOVA does not account for the cyclical 
nature of MIRs and because this lack of adjustment is a 
noted source of concern for MIA studies based solely on 
this statistical test (Okamura et al., 2013), a circular- 
linear model (see Okamura et al., 2013) was fit to the 
data from this study to analyze how many cycles (i.e., 
annuli) exist in a time span of 1 year. This method was 
used to assess AIC values for 3 models: models with no 
cycle (model N), 1 cycle (model A), and 2 cycles (model B) 
in the MIR data. This method was used separately for 
each age bin as well as for each region (with age bins 
combined). This analysis was completed in R, by using 
