<s>
There	O
are	O
two	O
main	O
uses	O
of	O
the	O
term	O
calibration	B-General_Concept
in	O
statistics	O
that	O
denote	O
special	O
types	O
of	O
statistical	O
inference	O
problems	O
.	O
</s>
<s>
a	O
reverse	O
process	O
to	O
regression	B-General_Concept
,	O
where	O
instead	O
of	O
a	O
future	O
dependent	O
variable	O
being	O
predicted	O
from	O
known	O
explanatory	O
variables	O
,	O
a	O
known	O
observation	O
of	O
the	O
dependent	O
variables	O
is	O
used	O
to	O
predict	O
a	O
corresponding	O
explanatory	O
variable	O
;	O
</s>
<s>
procedures	O
in	O
statistical	B-General_Concept
classification	I-General_Concept
to	O
determine	O
class	B-General_Concept
membership	I-General_Concept
probabilities	I-General_Concept
which	O
assess	O
the	O
uncertainty	O
of	O
a	O
given	O
new	O
observation	O
belonging	O
to	O
each	O
of	O
the	O
already	O
established	O
classes	O
.	O
</s>
<s>
In	O
addition	O
,	O
"	O
calibration	B-General_Concept
"	O
is	O
used	O
in	O
statistics	O
with	O
the	O
usual	O
general	O
meaning	O
of	O
calibration	B-General_Concept
.	O
</s>
<s>
For	O
example	O
,	O
model	O
calibration	B-General_Concept
can	O
be	O
also	O
used	O
to	O
refer	O
to	O
Bayesian	O
inference	O
about	O
the	O
value	O
of	O
a	O
model	O
's	O
parameters	O
,	O
given	O
some	O
data	O
set	O
,	O
or	O
more	O
generally	O
to	O
any	O
type	O
of	O
fitting	O
of	O
a	O
statistical	O
model	O
.	O
</s>
<s>
The	O
calibration	B-General_Concept
problem	O
in	O
regression	B-General_Concept
is	O
the	O
use	O
of	O
known	O
data	O
on	O
the	O
observed	O
relationship	O
between	O
a	O
dependent	O
variable	O
and	O
an	O
independent	O
variable	O
to	O
make	O
estimates	O
of	O
other	O
values	O
of	O
the	O
independent	O
variable	O
from	O
new	O
observations	O
of	O
the	O
dependent	O
variable	O
.	O
</s>
<s>
This	O
can	O
be	O
known	O
as	O
"	O
inverse	O
regression	B-General_Concept
"	O
:	O
see	O
also	O
sliced	B-Algorithm
inverse	I-Algorithm
regression	I-Algorithm
.	O
</s>
<s>
Calibration	B-General_Concept
in	O
classification	B-General_Concept
means	O
turning	O
transform	O
classifier	B-General_Concept
scores	O
into	O
class	B-General_Concept
membership	I-General_Concept
probabilities	I-General_Concept
.	O
</s>
<s>
The	O
following	O
univariate	O
calibration	B-General_Concept
methods	O
exist	O
for	O
transforming	O
classifier	B-General_Concept
scores	O
into	O
class	B-General_Concept
membership	I-General_Concept
probabilities	I-General_Concept
in	O
the	O
two-class	B-General_Concept
case	O
:	O
</s>
<s>
The	O
following	O
multivariate	O
calibration	B-General_Concept
methods	O
exist	O
for	O
transforming	O
classifier	B-General_Concept
scores	O
into	O
class	B-General_Concept
membership	I-General_Concept
probabilities	I-General_Concept
in	O
the	O
case	O
with	O
classes	O
count	O
greater	O
than	O
two	O
:	O
</s>
<s>
Philip	O
E	O
.	O
Tetlock	O
employs	O
the	O
term	O
"	O
calibration	B-General_Concept
"	O
in	O
this	O
sense	O
in	O
his	O
2015	O
book	O
Superforecasting	O
.	O
</s>
<s>
For	O
example	O
,	O
as	O
expressed	O
by	O
Daniel	O
Kahneman	O
,	O
"	O
if	O
you	O
give	O
all	O
events	O
that	O
happen	O
a	O
probability	O
of	O
.6	O
and	O
all	O
the	O
events	O
that	O
do	O
n’t	O
happen	O
a	O
probability	O
of	O
.4	O
,	O
your	O
discrimination	O
is	O
perfect	O
but	O
your	O
calibration	B-General_Concept
is	O
miserable	O
"	O
.	O
</s>
