<s>
Soft	O
independent	O
modelling	O
by	O
class	O
analogy	O
(	O
SIMCA	O
)	O
is	O
a	O
statistical	O
method	O
for	O
supervised	B-General_Concept
classification	I-General_Concept
of	O
data	O
.	O
</s>
<s>
In	O
order	O
to	O
build	O
the	O
classification	O
models	O
,	O
the	O
samples	O
belonging	O
to	O
each	O
class	O
need	O
to	O
be	O
analysed	O
using	O
principal	B-Application
component	I-Application
analysis	I-Application
(	O
PCA	O
)	O
;	O
only	O
the	O
significant	O
components	O
are	O
retained	O
.	O
</s>
<s>
For	O
a	O
given	O
class	O
,	O
the	O
resulting	O
model	O
then	O
describes	O
either	O
a	O
line	O
(	O
for	O
one	O
Principal	B-Application
Component	I-Application
or	O
PC	O
)	O
,	O
plane	O
(	O
for	O
two	O
PCs	O
)	O
or	O
hyper-plane	O
(	O
for	O
more	O
than	O
two	O
PCs	O
)	O
.	O
</s>
<s>
This	O
critical	O
distance	O
is	O
based	O
on	O
the	O
F-distribution	B-General_Concept
and	O
is	O
usually	O
calculated	O
using	O
95%	O
or	O
99%	O
confidence	O
intervals	O
.	O
</s>
<s>
The	O
classification	O
efficiency	O
is	O
usually	O
indicated	O
by	O
Receiver	B-Algorithm
operating	I-Algorithm
characteristics	I-Algorithm
.	O
</s>
<s>
In	O
the	O
original	O
SIMCA	O
method	O
,	O
the	O
ends	O
of	O
the	O
hyper-plane	O
of	O
each	O
class	O
are	O
closed	O
off	O
by	O
setting	O
statistical	O
control	O
limits	O
along	O
the	O
retained	O
principal	B-Application
components	I-Application
axes	O
(	O
i.e.	O
,	O
score	O
value	O
between	O
plus	O
and	O
minus	O
0.5	O
times	O
score	O
standard	O
deviation	O
)	O
.	O
</s>
