<s>
In	O
statistics	O
,	O
exploratory	B-General_Concept
data	I-General_Concept
analysis	I-General_Concept
(	O
EDA	O
)	O
is	O
an	O
approach	O
of	O
analyzing	B-General_Concept
data	B-General_Concept
sets	I-General_Concept
to	O
summarize	O
their	O
main	O
characteristics	O
,	O
often	O
using	O
statistical	B-Application
graphics	I-Application
and	O
other	O
data	B-Application
visualization	I-Application
methods	O
.	O
</s>
<s>
Exploratory	B-General_Concept
data	I-General_Concept
analysis	I-General_Concept
has	O
been	O
promoted	O
by	O
John	O
Tukey	O
since	O
1970	O
to	O
encourage	O
statisticians	O
to	O
explore	O
the	O
data	O
,	O
and	O
possibly	O
formulate	O
hypotheses	O
that	O
could	O
lead	O
to	O
new	O
data	O
collection	O
and	O
experiments	O
.	O
</s>
<s>
EDA	O
is	O
different	O
from	O
initial	O
data	B-General_Concept
analysis	I-General_Concept
(	O
IDA	O
)	O
,	O
which	O
focuses	O
more	O
narrowly	O
on	O
checking	O
assumptions	O
required	O
for	O
model	O
fitting	O
and	O
hypothesis	O
testing	O
,	O
and	O
handling	O
missing	O
values	O
and	O
making	O
transformations	O
of	O
variables	O
as	O
needed	O
.	O
</s>
<s>
Tukey	O
defined	O
data	B-General_Concept
analysis	I-General_Concept
in	O
1961	O
as	O
:	O
"	O
Procedures	O
for	O
analyzing	B-General_Concept
data	O
,	O
techniques	O
for	O
interpreting	O
the	O
results	O
of	O
such	O
procedures	O
,	O
ways	O
of	O
planning	O
the	O
gathering	O
of	O
data	O
to	O
make	O
its	O
analysis	O
easier	O
,	O
more	O
precise	O
or	O
more	O
accurate	O
,	O
and	O
all	O
the	O
machinery	O
and	O
results	O
of	O
(	O
mathematical	O
)	O
statistics	O
which	O
apply	O
to	O
analyzing	B-General_Concept
data.	O
"	O
</s>
<s>
Tukey	O
's	O
championing	O
of	O
EDA	O
encouraged	O
the	O
development	O
of	O
statistical	B-Algorithm
computing	I-Algorithm
packages	O
,	O
especially	O
S	B-Language
at	O
Bell	O
Labs	O
.	O
</s>
<s>
The	O
S	B-Language
programming	I-Language
language	I-Language
inspired	O
the	O
systems	O
S-PLUS	B-Application
and	O
R	B-Language
.	O
This	O
family	O
of	O
statistical-computing	O
environments	O
featured	O
vastly	O
improved	O
dynamic	O
visualization	O
capabilities	O
,	O
which	O
allowed	O
statisticians	O
to	O
identify	O
outliers	O
,	O
trends	O
and	O
patterns	O
in	O
data	O
that	O
merited	O
further	O
study	O
.	O
</s>
<s>
Tukey	O
's	O
EDA	O
was	O
related	O
to	O
two	O
other	O
developments	O
in	O
statistical	O
theory	O
:	O
robust	O
statistics	O
and	O
nonparametric	B-General_Concept
statistics	I-General_Concept
,	O
both	O
of	O
which	O
tried	O
to	O
reduce	O
the	O
sensitivity	O
of	O
statistical	O
inferences	O
to	O
errors	O
in	O
formulating	O
statistical	O
models	O
.	O
</s>
<s>
Tukey	O
promoted	O
the	O
use	O
of	O
five	B-Language
number	I-Language
summary	I-Language
of	O
numerical	O
data	O
—	O
the	O
two	O
extremes	O
(	O
maximum	O
and	O
minimum	O
)	O
,	O
the	O
median	O
,	O
and	O
the	O
quartiles	O
—	O
because	O
these	O
median	O
and	O
quartiles	O
,	O
being	O
functions	O
of	O
the	O
empirical	B-General_Concept
distribution	I-General_Concept
are	O
defined	O
for	O
all	O
distributions	O
,	O
unlike	O
the	O
mean	O
and	O
standard	B-General_Concept
deviation	I-General_Concept
;	O
moreover	O
,	O
the	O
quartiles	O
and	O
median	O
are	O
more	O
robust	O
to	O
skewed	B-General_Concept
or	O
heavy-tailed	O
distributions	O
than	O
traditional	O
summaries	O
(	O
the	O
mean	O
and	O
standard	B-General_Concept
deviation	I-General_Concept
)	O
.	O
</s>
<s>
The	O
packages	O
S	B-Language
,	O
S-PLUS	B-Application
,	O
and	O
R	B-Language
included	O
routines	O
using	O
resampling	B-General_Concept
statistics	I-General_Concept
,	O
such	O
as	O
Quenouille	O
and	O
Tukey	O
's	O
jackknife	O
and	O
Efron	O
bootstrap	B-Application
,	O
which	O
are	O
nonparametric	B-General_Concept
and	O
robust	O
(	O
for	O
many	O
problems	O
)	O
.	O
</s>
<s>
Exploratory	B-General_Concept
data	I-General_Concept
analysis	I-General_Concept
,	O
robust	O
statistics	O
,	O
nonparametric	B-General_Concept
statistics	I-General_Concept
,	O
and	O
the	O
development	O
of	O
statistical	O
programming	O
languages	O
facilitated	O
statisticians	O
 '	O
work	O
on	O
scientific	O
and	O
engineering	O
problems	O
.	O
</s>
<s>
These	O
statistical	O
developments	O
,	O
all	O
championed	O
by	O
Tukey	O
,	O
were	O
designed	O
to	O
complement	O
the	O
analytic	B-Language
theory	O
of	O
testing	O
statistical	O
hypotheses	O
,	O
particularly	O
the	O
Laplacian	O
tradition	O
's	O
emphasis	O
on	O
exponential	O
families	O
.	O
</s>
<s>
John	O
W	O
.	O
Tukey	O
wrote	O
the	O
book	O
Exploratory	B-General_Concept
Data	I-General_Concept
Analysis	I-General_Concept
in	O
1977	O
.	O
</s>
<s>
Tukey	O
held	O
that	O
too	O
much	O
emphasis	O
in	O
statistics	O
was	O
placed	O
on	O
statistical	O
hypothesis	O
testing	O
(	O
confirmatory	O
data	B-General_Concept
analysis	I-General_Concept
)	O
;	O
more	O
emphasis	O
needed	O
to	O
be	O
placed	O
on	O
using	O
data	O
to	O
suggest	O
hypotheses	O
to	O
test	O
.	O
</s>
<s>
In	O
particular	O
,	O
he	O
held	O
that	O
confusing	O
the	O
two	O
types	O
of	O
analyses	O
and	O
employing	O
them	O
on	O
the	O
same	O
set	B-General_Concept
of	I-General_Concept
data	I-General_Concept
can	O
lead	O
to	O
systematic	O
bias	O
owing	O
to	O
the	O
issues	O
inherent	O
in	O
testing	O
hypotheses	O
suggested	O
by	O
the	O
data	O
.	O
</s>
<s>
Many	O
EDA	O
techniques	O
have	O
been	O
adopted	O
into	O
data	B-Application
mining	I-Application
.	O
</s>
<s>
Typical	O
graphical	B-Application
techniques	I-Application
used	O
in	O
EDA	O
are	O
:	O
</s>
<s>
Dimensionality	B-Algorithm
reduction	I-Algorithm
:	O
</s>
<s>
Francis	O
Galton	O
emphasized	O
order	B-General_Concept
statistics	I-General_Concept
and	O
quantiles	O
.	O
</s>
<s>
Arthur	O
Lyon	O
Bowley	O
used	O
precursors	O
of	O
the	O
stemplot	B-Application
and	O
five-number	B-Language
summary	I-Language
(	O
Bowley	O
actually	O
used	O
a	O
"	O
seven-figure	O
summary	O
"	O
,	O
including	O
the	O
extremes	O
,	O
deciles	O
and	O
quartiles	O
,	O
along	O
with	O
the	O
mediansee	O
his	O
Elementary	O
Manual	O
of	O
Statistics	O
(	O
3rd	O
edn.	O
,	O
1920	O
)	O
,	O
p.62	O
–	O
he	O
defines	O
"	O
the	O
maximum	O
and	O
minimum	O
,	O
median	O
,	O
quartiles	O
and	O
two	O
deciles	O
"	O
as	O
the	O
"	O
seven	O
positions	O
"	O
)	O
.	O
</s>
<s>
Andrew	O
Ehrenberg	O
articulated	O
a	O
philosophy	O
of	O
data	B-General_Concept
reduction	I-General_Concept
(	O
see	O
his	O
book	O
of	O
the	O
same	O
name	O
)	O
.	O
</s>
<s>
The	O
Open	O
University	O
course	O
Statistics	O
in	O
Society	O
(	O
MDST	O
242	O
)	O
,	O
took	O
the	O
above	O
ideas	O
and	O
merged	O
them	O
with	O
Gottfried	O
Noether	O
's	O
work	O
,	O
which	O
introduced	O
statistical	O
inference	O
via	O
coin-tossing	O
and	O
the	O
median	B-General_Concept
test	I-General_Concept
.	O
</s>
<s>
JMP	B-Language
,	O
an	O
EDA	O
package	O
from	O
SAS	O
Institute	O
.	O
</s>
<s>
KNIME	B-Language
,	O
Konstanz	O
Information	O
Miner	O
–	O
Open-Source	B-Application
data	O
exploration	O
platform	O
based	O
on	O
Eclipse	O
.	O
</s>
<s>
Minitab	B-Application
,	O
an	O
EDA	O
and	O
general	O
statistics	O
package	O
widely	O
used	O
in	O
industrial	O
and	O
corporate	O
settings	O
.	O
</s>
<s>
Orange	B-Application
,	O
an	O
open-source	B-Application
data	B-Application
mining	I-Application
and	O
machine	O
learning	O
software	O
suite	O
.	O
</s>
<s>
Python	B-Language
,	O
an	O
open-source	B-Application
programming	O
language	O
widely	O
used	O
in	O
data	B-Application
mining	I-Application
and	O
machine	O
learning	O
.	O
</s>
<s>
R	B-Language
,	O
an	O
open-source	B-Application
programming	O
language	O
for	O
statistical	B-Algorithm
computing	I-Algorithm
and	O
graphics	O
.	O
</s>
<s>
Together	O
with	O
Python	B-Language
one	O
of	O
the	O
most	O
popular	O
languages	O
for	O
data	O
science	O
.	O
</s>
<s>
TinkerPlots	B-Application
an	O
EDA	O
software	O
for	O
upper	O
elementary	O
and	O
middle	O
school	O
students	O
.	O
</s>
<s>
Weka	B-Language
an	O
open	O
source	O
data	B-Application
mining	I-Application
package	O
that	O
includes	O
visualization	O
and	O
EDA	O
tools	O
such	O
as	O
targeted	B-Application
projection	I-Application
pursuit	I-Application
.	O
</s>
