<s>
Twisting	B-Algorithm
properties	I-Algorithm
in	O
general	O
terms	O
are	O
associated	O
with	O
the	O
properties	O
of	O
samples	O
that	O
identify	O
with	O
statistics	O
that	O
are	O
suitable	O
for	O
exchange	O
.	O
</s>
<s>
Starting	O
with	O
a	O
sample	O
observed	O
from	O
a	O
random	O
variable	O
X	O
having	O
a	O
given	O
distribution	O
law	O
with	O
a	O
non-set	O
parameter	O
,	O
a	O
parametric	B-General_Concept
inference	I-General_Concept
problem	O
consists	O
of	O
computing	O
suitable	O
values	O
–	O
call	O
them	O
estimates	O
–	O
of	O
this	O
parameter	O
precisely	O
on	O
the	O
basis	O
of	O
the	O
sample	O
.	O
</s>
<s>
In	O
algorithmic	B-General_Concept
inference	I-General_Concept
,	O
suitability	O
of	O
an	O
estimate	O
reads	O
in	O
terms	O
of	O
compatibility	O
with	O
the	O
observed	O
sample	O
.	O
</s>
<s>
In	O
more	O
abstract	O
terms	O
,	O
we	O
speak	O
about	O
twisting	B-Algorithm
properties	I-Algorithm
of	O
samples	O
with	O
properties	O
of	O
parameters	O
and	O
identify	O
the	O
former	O
with	O
statistics	O
that	O
are	O
suitable	O
for	O
this	O
exchange	O
,	O
so	O
denoting	O
a	O
well	B-General_Concept
behavior	I-General_Concept
w.r.t.	O
</s>
<s>
When	O
s	O
is	O
a	O
well-behaved	B-General_Concept
statistic	I-General_Concept
w.r.t	O
the	O
parameter	O
,	O
we	O
are	O
sure	O
that	O
a	O
monotone	O
relation	O
exists	O
for	O
each	O
between	O
s	O
and	O
θ	O
.	O
</s>
<s>
Identify	O
a	O
well	B-General_Concept
behaving	I-General_Concept
statistic	I-General_Concept
S	O
for	O
the	O
parameter	O
θ	O
and	O
its	O
discretization	O
grain	O
(	O
if	O
any	O
)	O
;	O
</s>
<s>
Instead	O
,	O
the	O
difficulty	O
of	O
dealing	O
with	O
a	O
vector	O
of	O
parameters	O
proved	O
to	O
be	O
the	O
Achilles	O
heel	O
of	O
Fisher	O
's	O
approach	O
to	O
the	O
fiducial	B-Error_Name
distribution	I-Error_Name
of	O
parameters	O
.	O
</s>
