Jacobson et at: Measurement errors in body size of Placopecten magellanicus 
237 
for distributions of measurement errors because 
they were closer to normally distributed. 
Multiple shell height-measurements were usu- 
ally made from single specimens in our experi- 
ments. We made allowance for repeated sampling 
when testing skewness and kurtosis by using the 
number of unique specimens in the experiment as 
the degrees of freedom instead of the number of 
measurements (i.e., if n measurements were made 
on each of k specimens, we used k as the degrees 
of freedom in statistical tests). The effect of this 
adjustment was to make the statistical tests more 
conservative (less likely to reject the null hypoth- 
esis of no difference). The number of specimens 
is a reasonable lower bound estimate of the true 
effective sample size. 
Body weights for sea scallops and other marine 
organisms are often computed from body size. For 
sea scallops in this analysis, 
w = e a + pUh)^ ( 5 ) 
where W = sea scallop meat weight ( g , the weight 
of the marketable adductor muscle); 
h = shell height (mm); and the parameter 
values a=-12.01 and 3.22. 
Bland-Altman plots (1986, 1995) were used to 
characterize shell-height measurement errors. In 
the case of measuring boards, for example, the dif- 
ference between the measuring board and caliper 
shell-height measurements for each sea scallop was 
plotted on the y-axis against the average of the 
two measures for the same individual on the x- 
axis. Bland-Altman plots are typically presented 
as scatter plots with a point for each difference 
(pair of measurements); however, boxplots may be 
more useful in some circumstances (see below). 
Bland-Altman plots are useful because they elimi- 
nate spurious correlations when the difference of 
y-x is plotted against the more precise measure (x) 
and because patterns are easier to discern along a 
horizontal line (the x-axis) than along a diagonal 
line. Spurious correlations occur because the mea- 
surement error in x affects the variables plotted on 
both the x- and y-axes. 
Experiment 1 was designed to measure the accuracy 
of video measurements for objects of known size (square 
ceramic tiles) as a function of position in the video 
frame as measured by DFO (Fig. 2). Scuba divers in 
experiment 1 placed black and white ceramic floor tiles 
(all were 48.5x48.5 mm) in a closely packed square grid 
on the bottom of the tank, starting at the center of the 
video pyramid and covering the entire range of view 
in actual surveys (Fig. 2). The width and height of 91 
tiles across the field of view and at various distances 
and positions from the center of the sampling frame 
(Fig. 2) were estimated from video images by using the 
standard video survey procedures described above. Data 
were recorded in such a way that the length and height 
measurements from the same tile could be associated 
with each other and with the particular position of the 
tile in the video image. The tiles used in experiment 1 
(48.5x48.5 mm) corresponded roughly with the size of 
the smallest scallops fully recruited to the dredge and 
video surveys (about 40 mm SH) and included in stock 
assessment analyses. Sea scallops, according to actual 
survey data, cover a much wider range of shell heights 
(to about 190 mm SH in experiment 2, see Discussion 
section). 
Experiment 2 was designed to measure the accu- 
racy of video shell-height measurements for sea scal- 
lop shells of varying sizes (39 to 192 mm SH) placed 
randomly on a sand-granule-pebble substrate, similar 
