378
A. butcheri
15
~
:;;" 10
.!!
.!!
" f 5
Cl.
P. auratus
G
CI
•
15
" 10
~
Ü
:;;
•
5
Q.
5
10
Oblerved age
5
10
Oblerved age
M. novaezelandiae
·
CI
G
15
" 10
~
:;;
f
5
Cl.
5
10
Oblerved age
15
15
15
S.G. Robertson . A.K. Morison
50
40
>~ 30
• ~
l 20
u.
>10
40
35
30
~ 25
~ 20
: 15
u: 10
5
100
80
>u
60
~
~
~
.,. 40
.
,::
20
o
- -6 - Predlcled
10
15
Age
--+-Ob •• rved
- -6 - Predlcled
10
Age
15
-+-Oblerved
• -6 - Predlcled
10
Age
15
Figure 19.5. Age bias (left) and age composition (right) plots for each species.
The age bias plots show mean (± 2 SE) predicted ages against observed age; the
line shows equal age estimates.
19.6
Discussion
These results confirm the findings from the preliminary study of Robertson and
Morison (1999) that neural networks offer a practical means to objectively
estimate the age of fish. More confidence can be placed in the results presented
here because of the larger sampie sizes and the more rigorous testing procedures.