<s>
Collostructional	B-General_Concept
analysis	I-General_Concept
is	O
a	O
family	O
of	O
methods	O
developed	O
by	O
(	O
in	O
alphabetical	O
order	O
)	O
Stefan	O
Th	O
.	O
</s>
<s>
Collostructional	B-General_Concept
analysis	I-General_Concept
aims	O
at	O
measuring	O
the	O
degree	O
of	O
attraction	O
or	O
repulsion	O
that	O
words	O
exhibit	O
to	O
constructions	O
,	O
where	O
the	O
notion	O
of	O
construction	O
has	O
so	O
far	O
been	O
that	O
of	O
Goldberg	O
's	O
construction	O
grammar	O
.	O
</s>
<s>
Collostructional	B-General_Concept
analysis	I-General_Concept
so	O
far	O
comprises	O
three	O
different	O
methods	O
:	O
</s>
<s>
Collostructional	B-General_Concept
analysis	I-General_Concept
requires	O
frequencies	O
of	O
words	O
and	O
constructions	O
and	O
is	O
similar	O
to	O
a	O
wide	O
variety	O
of	O
collocation	B-General_Concept
statistics	O
.	O
</s>
<s>
(	O
i	O
)	O
a	O
comparison	O
of	O
the	O
observed	O
frequency	O
to	O
the	O
one	O
expected	O
by	O
chance	O
;	O
thus	O
,	O
collostructional	B-General_Concept
analysis	I-General_Concept
can	O
distinguish	O
attraction	O
and	O
repulsion	O
of	O
words	O
and	O
constructions	O
;	O
</s>
<s>
(	O
ii	O
)	O
a	O
measure	O
of	O
the	O
strength	O
of	O
the	O
attraction	O
or	O
repulsion	O
;	O
this	O
is	O
usually	O
the	O
log-transformed	O
p-value	B-General_Concept
of	O
a	O
Fisher-Yates	B-General_Concept
exact	I-General_Concept
test	I-General_Concept
.	O
</s>
<s>
(	O
i	O
)	O
it	O
measures	O
not	O
the	O
association	O
of	O
words	O
to	O
words	O
,	O
but	O
of	O
words	O
to	O
syntactic	O
patterns	O
or	O
constructions	O
;	O
thus	O
,	O
it	O
takes	O
syntactic	O
structure	O
more	O
seriously	O
than	O
most	O
collocation-based	O
analyses	O
;	O
</s>
<s>
(	O
ii	O
)	O
it	O
has	O
so	O
far	O
only	O
used	O
the	O
most	O
precise	O
statistics	O
,	O
namely	O
the	O
Fisher-Yates	B-General_Concept
exact	I-General_Concept
test	I-General_Concept
based	O
on	O
the	O
hypergeometric	O
distribution	O
;	O
thus	O
,	O
unlike	O
t-scores	O
,	O
z-scores	O
,	O
chi-square	B-General_Concept
tests	I-General_Concept
etc.	O
,	O
the	O
analysis	O
is	O
not	O
based	O
on	O
,	O
and	O
does	O
not	O
violate	O
,	O
any	O
distributional	O
assumptions	O
.	O
</s>
