Walt
Stevenson/
RTP/
USEPA/
US
04/
26/
2006
05:
57
PM
To
etoy@
omb.
eop.
gov
cc
Rick
Vetter/
RTP/
USEPA/
US@
EPA,
RobertJ
Wayland/
RTP/
USEPA/
US@
EPA,
Brian
Shrager/
RTP/
USEPA/
US@
EPA
Subject
LMWC
Cost
text
Edmond
Here
is
revised
text
for
the
cost
issue.
This
text
is
satisfactory
to
EPA.
Do
we
have
any
remaining
issues
??

Walt
IV.
Impacts
of
the
Final
Amendments
for
Existing
Units
The
EPA
projects
the
final
amendments
will
have
no
additional
impacts
to
air,
water,
or
energy
since
the
final
emission
limits
can
be
achieved
using
the
same
air
pollution
control
technology
that
was
used
to
comply
with
the
current
emission
limits.
Similarly,
EPA
expects
minimal
cost
and
no
economic
impact
for
the
same
reason.
Existing
large
MWC
units
will
continue
to
use
their
existing
MACT
control
technology
to
meet
the
emission
limits,
and
will
not
incur
costs
to
retrofit
equipment.
In
addition,
EPA
does
not
believe
that
the
revised
limits
will
result
in
any
increase
in
operating
or
maintenance
costs.
The
same
conclusions
apply
to
new
MWC
units
since
EPA
expects
that
new
MWC
units
will
be
equipped
with
the
same
control
technology
used
to
comply
with
the
1995
NSPS.
Statutory
and
Executive
Order
Reviews
­­­­­
Forwarded
by
Brian
Shrager/
RTP/
USEPA/
US
on
04/
27/
2006
01:
03
PM
­­­­­

Walt
Stevenson/
RTP/
USEPA/
US
04/
26/
2006
06:
15
PM
To
"
Toy,
Edmond"
<
Edmond_
Toy@
omb.
eop.
gov>
cc
Brian
Shrager/
RTP/
USEPA/
US@
EPA,
Rick
Vetter/
RTP/
USEPA/
US@
EPA,
RobertJ
Wayland/
RTP/
USEPA/
US@
EPA
Sub
ject
RE:
Internet
notes
on
percentile
and
UCL
Edmond
Lets
wait
for
now.
I
have
already
asked
the
statistician
to
prepare
a
note
on
the
subject.
We
should
have
it
later
tonight..
I'll
forward
it
to
you
when
I
receive
it.
I'll
call
them
in
the
AM
if
I
do
not
receive
the
note
this
evening
Walt
"
Toy,
Edmond"
<
Edmond_
Toy@
omb.
eop.
gov>

04/
26/
2006
06:
06
PM
To
Walt
Stevenson/
RTP/
USEPA/
US@
EPA
cc
RobertJ
Wayland/
RTP/
USEPA/
US@
EPA,
Rick
Vetter/
RTP/
USEPA/
US@
EPA,
Brian
Shrager/
RTP/
USEPA/
US@
EPA
Sub
ject
RE:
Internet
notes
on
percentile
and
UCL
I
still
think
this
use
of
terminology
is
very
unconventional
and
should
be
changed.
Should
I
talk
to
your
statistician?

­­­­­
Original
Message­­­­­
From:
Stevenson.
Walt@
epamail.
epa.
gov
[
mailto:
Stevenson.
Walt@
epamail.
epa.
gov]
Sent:
Wednesday,
April
26,
2006
4:
57
PM
To:
Toy,
Edmond
Cc:
Wayland.
Robertj@
epamail.
epa.
gov;
Vetter.
Rick@
epamail.
epa.
gov;
Shrager.
Brian@
epamail.
epa.
gov
Subject:
Fw:
Internet
notes
on
percentile
and
UCL
Edmond
FYI
Walt
Percentile
(
from
Wikipedia):

In
descriptive
statistics,
a
percentile
is
any
of
the
99
values
that
divide
the
sorted
data
into
100
equal
parts,
so
that
each
part
represents
1/
100th
(
or
1%)
of
the
sample
or
population.

Confidence
interval
(
or
limits):
(
also
from
Wikipedia)

In
statistics,
a
confidence
interval
(
CI)
is
an
interval
between
two
numbers,
where
there
is
a
certain
specified
level
of
confidence
that
a
population
parameter
lies.
Confidence
intervals
are
the
most
prevalent
form
of
interval
estimation.

Walt,
note
that
when
we
calculated
UCL
values,
we
took
several
steps
in
getting
there,
the
larger
of
which
being:

1.
Determining
the
distribution
type
and
transforming
to
get
a
normal
transformation.

2.
Applying
t­
values
to
determine
the
confidence
limit.

If,
however,
we
were
calculating
a
percentile,
we
would
find
the
range
of
the
datapoints,
sort
the
data
and
then
divide
by
100
to
get
the
value
that
would
separate
the
range
into
100
equal
parts.
No
population
distribution
tests,
transformations,
or
t­
or
z­
tests
of
probability
are
required
for
the
percentile
calculation.
"
Toy,
Edmond"
<
Edmond_
Toy@
omb.
eop.
gov>

04/
27/
2006
09:
45
AM
To
Walt
Stevenson/
RTP/
USEPA/
US@
EPA
cc
Brian
Shrager/
RTP/
USEPA/
US@
EPA,
Rick
Vetter/
RTP/
USEPA/
US@
EPA,
RobertJ
Wayland/
RTP/
USEPA/
US@
EPA
Subject
RE:
Internet
notes
on
percentile
and
UCL
Walt...

Please
pass
this
on
to
your
statistician.

The
ERG
memo
that
you
sent
me
on
the
statistical
methodology
talks
about
the
UCL
(
page
5,
equation
5,
and
footnote
3).
The
footnote
states
that
the
term
"
UCL"
does
NOT
refer
to
the
upper
confidence
limit
one
would
compute
for
a
population
mean
based
on
a
sample,
and
the
footnote
goes
on
to
give
the
equation
that
every
textbook
uses
for
UCL.
The
footnote
equation
is
different
than
eq.
5
in
the
body
of
the
memo,
which
is
what
ERG
calls
a
"
UCL".
I
think
this
footnote
acknowledges
that
what
ERG
is
calling
a
"
UCL"
is
an
unconventional
use
of
the
term.

Every
textbook
that
I
can
find
defines
the
UCL
as
the
mean
plus/
minus
the
(
critical
t
value
times
the
standard
error
of
the
estimate)...
and
that
is
what
is
in
the
footnote.
However,
eq
5
doesn't
use
the
standard
error
of
the
estimate,
but
rather
the
standard
devation
of
the
whole
distribution
(
in
our
case,
the
emissions
rate).
Eq
5
looks
almost
identical
to
the
equation
for
percentile.

Thinking
in
words,
EPA
wants
to
select
as
the
standard
the
emissions
rate
that
is
expected
to
be
met
99%
or
99.7%
of
the
time.
This
is
conceptually
identical
to
the
notion
of
the
99th
percentile:
if
you
take
a
sample
from
your
distribution,
99%
of
your
observations
would
be
expected
to
less
than
or
equal
to
the
99th
percentile.

­
Edmond
"
Toy,
Edmond"
<
Edmond_
Toy@
omb.
eop.
gov>

04/
27/
2006
09:
57
AM
To
Walt
Stevenson/
RTP/
USEPA/
US@
EPA
cc
Rick
Vetter/
RTP/
USEPA/
US@
EPA,
RobertJ
Wayland/
RTP/
USEPA/
US@
EPA,
Brian
Shrager/
RTP/
USEPA/
US@
EPA
Subject
RE:
LMWC
Cost
text
Please
change
the
heading
of
the
section
to
delete
"
for
Existing
Units"
(
the
discussion
is
not
limited
to
existing
units)

Walt
Stevenson/
RTP/
USEPA/
US
04/
27/
2006
11:
01
AM
To
"
Toy,
Edmond"
<
Edmond_
Toy@
omb.
eop.
gov>
cc
Brian
Shrager/
RTP/
USEPA/
US@
EPA,
Rick
Vetter/
RTP/
USEPA/
US@
EPA,
RobertJ
Wayland/
RTP/
USEPA/
US@
EPA
Subject
RE:
LMWC
Cost
text
will
do
Walt
Stevenson/
RTP/
USEPA/
US
04/
27/
2006
11:
00
AM
To
"
Toy,
Edmond"
<
Edmond_
Toy@
omb.
eop.
gov>
cc
Brian
Shrager/
RTP/
USEPA/
US@
EPA,
Rick
Vetter/
RTP/
USEPA/
US@
EPA,
RobertJ
Wayland/
RTP/
USEPA/
US@
EPA
Subject
RE:
Internet
notes
on
percentile
and
UCL
Edmond
I
will
change
the
preamble
discussion
to
use
the
term"
percentile"
in
place
of
the
term
"
upper
confidence
limit"

Walt
