Elicitation Protocol
Page 1 of 21
Protocol for Probabilistic Characterization of
Uncertainty in Mortality Response to Airborne
Fine Particulate Matter
Elicitation of European Air Pollution Experts
Harvard-Kuwait Public Health Project
Version 5-3, Printed: June 16, 2004
Elicitation Protocol
Part I:
Page 2 of 21
Introduction
Thank you for participating in this expert judgment study of the concentration-response
relationship between airborne fine particulate matter and population mortality rates.
Objectives
There are two main objectives of this study:
1. to probabilistically characterize the concentration-response relationship between
airborne fine particulate matter (PM2.5) and population mortality rates.
2. to gain qualitative insight into factors influencing the PM2.5 mortality
concentration-response relationship.
Method
The method employed is structured expert judgment. Experts’ uncertainty estimates for
variables of interest (related to PM2.5 mortality concentration-response) are elicited via a
structured protocol of questions. Included in the elicitation are variables whose true
values will become known in the time frame of the study. These variables are used to
measure and hopefully validate the expert performance uncertainty quantifications. A
good uncertainty assessor is one who

gives assessments which are statistically accurate, and

gives assessments which are informative.
Statistical accuracy means roughly the following. For each of a number of uncertain
quantities the expert gives an interval which he is 90% certain will contain the
(unknown) true value. Then the assessments are statistically accurate if (in an
appropriate statistical sense) 90% of the true values fall within the respective
intervals. A similar definition can be applied for other probability intervals.
Informativeness means roughly that the probability intervals are concentrated, e.g.
that the assessor is 90% certain of catching the true values in small intervals.
Both performance indicators are important, but statistical accuracy is arguably more
important than informativeness. Expert assessments may be combined in a number of
different ways, including equal weighting or by one of several approaches designed to
give differential weights to each expert. In a performance-based approach, differential
weights are given to experts according to performance indicators.
Version 5-3, Printed: June 16, 2004
Elicitation Protocol
Page 3 of 21
Confidentiality/feedback
Following the individual elicitation, we will summarize your responses in free text
format. Each expert will receive a copy of his/her quantitative and qualitative
assessments for review and will receive information about how the assessments will be
used. The names and qualifications of the experts will be published in the final report, as
will the individual expert assessments and all information relevant. The names of the
experts will not be directly identified with individual assessments in the published
reports. However, this information will be preserved with the project files and may be
made available to in the case of an authorized peer review.
Elicitation Format
We ask you to probabilistically characterize the PM2.5 concentration-response in several
different settings representing variations in:




particle composition,
baseline pollution conditions,
age-structure and background health status of the exposed population, and
exposure duration.
We will ask you eight questions about the mortality concentration-response relationship
in different settings. These are termed “variables of interest” for this elicitation project.
In addition to these eight questions, we will ask you supplemental questions about the
intersection of short-term and long-term exposure studies and about the potential for
different particle constituents to have different toxicities. Throughout these questions, we
will ask you to describe your reasoning for the uncertainty estimates that you provide.
Following our elicitation of your uncertainty estimates for these variables, we will ask
you, in a similar fashion, to answer several more questions relevant to the variables of
interest. These are termed “seed questions” and they are designed to gauge the
calibration and informativeness of each expert. These questions differ from variables of
interest in that their answers are observable quantities. Therefore, performance on these
questions may be assessed and used to determine expert weights, as previously explained.
Version 5-3, Printed: June 16, 2004
Elicitation Protocol
Page 4 of 21
Variables of Interest
The questions for the seven variables of interest share a common form, with slight
variations pertaining to different locations, exposure durations, and effect intervals for
each setting. For example, the first question is presented as follows:
What is your estimate of the true, but unknown, percent change in the total
annual, non-accidental mortality rate in the adult U.S. population
resulting from a permanent 1 μg/m3 reduction in long-term annual
average PM2.5 (from a population-weighted baseline concentration of 18
μg/m3) throughout the U.S.? To express the uncertainty associated with
the concentration-response relationship, please provide the 5th, 25th,
50th, 75th, and 95th percentiles of your estimate.
The underlined sections of text indicate elements of the general question which differ
from one specific question to another.
You are asked to express your uncertainty in terms of the 5th, 25th, 50th, 75th, and 95th
percentiles of your uncertainty distribution, where:



The 5th and 95th percentiles give an interval which you are 90% certain contains
the true value, with equal probability of falling above or below.
The 25th and 75th percentiles give an interval inside the 5th and 95th percentiles,
which you are 50% certain contains the true value, with equal probability of
falling above or below.
The 50th percentile, or median, is that value for which you think it is equally likely
that the true value fall above or below.
These notions may be represented graphically in terms of a probability density as shown
in the example question and hypothetical answer provided below. Because we do not
want to use an example that might bias your responses in this elicitation, the example
question is about benzene concentrations, rather than particulate matter concentrationresponse.
Example Question:
What is your estimate of the expected 6-day integrated benzene exposure
concentration in ambient air to be experienced by subjects in a forthcoming
exposure study (in ppm)? (Details of the study provided separately) To
express the uncertainty associated with this exposure concentration, please
provide the 5th, 25th, 50th, 75th, and 95th percentiles of your estimate.
Example Answer:
Elicited Values
Percentiles
1.5
0%
5
5%
Version 5-3, Printed: June 16, 2004
7
25%
8.8
50%
11
75%
14
95%
20
100%
Elicitation Protocol
Page 5 of 21
This table of elicited values corresponds to the following histogram:
25%
50%
0.14
75%
0.12
5%
0.1
0.08
95%
0.06
0.04
0.02
1.5
55
7
8.8
8.8
11
14
14
25
20
Note that although 0 and 100 percentiles are included in the example above, we will not
ask you to estimate these percentiles in this elicitation. We will ask you to provide only
the 5th, 25th, 50th, 75th, and 95th percentile estimates for each question.
The following grid represents the eight questions for the variables of interest and the
order in which they will be asked:
Exposure
duration
Time of
interest
for
effect
US
(Baseline:
18 µg/m3)
Permanent
Longterm
1
One day
1 week
3
One day
3 months
6
9 months
Longterm
Version 5-3, Printed: June 16, 2004
MCMA
(Baseline:
35 µg/m3)
Europe
(Baseline:
20 µg/m3)
Kuwait
(Specific
Exposure
Scenario)
2
4
5
11, 12
Elicitation Protocol
Page 6 of 21
Questions 7 – 10 are concerned with differences in toxicity among PM constituents and
intersection of short and long term exposures.
Definitions for each region of interest—including population profiles and baseline rates
of relevant variables—are contained in region-specific background sheets included in an
appendix to this protocol. 'US' in this context means the 48 states of the United States,
excluding Hawaii and Alaska. 'Europe' means the 15 countries that belonged to the
European Union before May 1st, 2004. 'MCMA' means Mexico City Metropolitan Area.
Rationale
Your opinions and interpretation of the evidence underlying these estimates are just as
important as your quantitative assessments. As we elicit your quantitative responses, we
will be discussing your reasoning and the evidence supporting your responses. The
qualitative questions below represent important issues underlying the quantitative
estimates that you give. They are meant to guide our discussion during the elicitation.
You will not be asked to write responses to each question; rather, we will ask your
opinion on these when they are relevant during the elicitation. Following the interview
we will prepare a summary of your responses and send them to you for review. You will
have 1 week to comment on these summaries before they are finalized for our project
report.
While we may not follow the questions below exactly, we want you to think about them
as a general guide and express your reasoning before you give us your quantitative
estimates. For each question, we would like you to incorporate the following questions
into your rationale:

What evidence suggests large values for this relationship?

What is the highest plausible value?

Tell us a little about your reasoning, the evidence, and theories that lead you to
this value.

Can you tell us of scenarios that would yield higher results?

What evidence or theory suggests small values?

What is the lowest plausible value?

Tell us a little about your reasoning, the evidence, and theories that lead you to
this value.

Can you tell us of scenarios that would yield lower results?
Version 5-3, Printed: June 16, 2004
Elicitation Protocol
Page 7 of 21
In a few minutes we will begin asking the elicitation questions about the variables of
interest. Before doing so, however, we would like to ask three general questions about
PM epidemiology results and causality:

What are the key properties of an ideal epidemiology study for measuring longterm mortality impacts of PM exposure?

Similarly, what key properties can you identify for studies of short-term mortality
impacts of PM exposure?

What factors need to be considered to decide whether epidemiology results should
be viewed as causal?
Version 5-3, Printed: June 16, 2004
Elicitation Protocol
Page 8 of 21
Part II: Elicitation of Uncertainty Distributions
A.
Variables of Interest
Question
1
Setting
US
Exposure
(Effect Interval)
Long-term
Change
1 μg/m3
Pollutant
PM2.5
Composition
Ambient
Baseline
18 ug/m3
What is your estimate of the true, but unknown, percent change in the total annual, non-accidental mortality
rate in the adult U.S. population resulting from a permanent 1 μg/m3 reduction in long-term annual average
PM2.5 (from a population-weighted baseline concentration of 18 μg/m3) throughout the U.S.? To express the
uncertainty associated with the concentration-response relationship, please provide the 5th, 25th, 50th, 75th, and
95th percentiles of your estimate.
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
Question
2
Setting
EU
Exposure
(Effect Interval)
Long-term
Change
1 μg/m3
Pollutant
PM2.5
Composition
Ambient
Baseline
20 ug/m3
What is your estimate of the true, but unknown, percent change in the total annual, non-accidental mortality
rate in the adult European population resulting from a permanent 1 μg/m3 reduction in long-term annual average
PM2.5 (from a population-weighted baseline concentration of 20 μg/m3) throughout the EU? To express the
uncertainty associated with the concentration-response relationship, please provide the 5th, 25th, 50th, 75th, and
95th percentiles of your estimate.
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
Version 5-3, Printed: June 16, 2004
Elicitation Protocol
Question
3
Setting
US
Page 9 of 21
Exposure
(Effect Interval)
Short-term (one week)
Change
10 μg/m3
Pollutant
PM2.5
Composition
Ambient
Baseline
Current
What is your estimate of the true, but unknown, percent change in non-accidental mortality in the total U.S.
population over the one week following a 10 μg/m3 increase in PM2.5 levels on a single day throughout the U.S.?
To express the uncertainty associated with the concentration-response relationship, please provide the 5th, 25th,
50th, 75th, and 95th percentiles of your estimate.
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
Question
4
Setting
MCMA
Exposure
(Effect Interval)
Short-term (one week)
Change
10 μg/m3
Pollutant
PM2.5
Composition
Ambient
Baseline
Current
What is your estimate of the true, but unknown, percent change in non-accidental mortality in the total
MCMA population over the one week following a 10 μg/m3 increase in PM2.5 levels on a single day throughout
the MCMA? To express the uncertainty associated with the concentration-response relationship, please
provide the 5th, 25th, 50th, 75th, and 95th percentiles of your estimate.
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
Question
5
Setting
EU
Exposure
(Effect Interval)
Short-term (one week)
Change
10 μg/m3
Pollutant
PM2.5
Composition
Ambient
Baseline
Current
What is your estimate of the true, but unknown, percent change in non-accidental mortality in the total
European population over the one week following a 10 μg/m3 increase in PM2.5 levels on a single day
throughout the EU? To express the uncertainty associated with the concentration-response relationship, please
provide the 5th, 25th, 50th, 75th, and 95th percentiles of your estimate.
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
Version 5-3, Printed: June 16, 2004
Elicitation Protocol
Question
6
Setting
US
Page 10 of 21
Exposure
(Effect Interval)
three months
Change
10 μg/m3
Pollutant
PM2.5
Composition
Ambient
Baseline
Current
What is your estimate of the true, but unknown, percent change in non-accidental mortality in the total U.S.
population over the three months following a 10 μg/m3 increase in PM2.5 levels on a single day throughout the
U.S.? To express the uncertainty associated with the concentration-response relationship, please provide the
5th, 25th, 50th, 75th, and 95th percentiles of your estimate.
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
Intersection of Short-term and Long-term Effects
One question that arises in the interpretation of cohort and time-series mortality studies is
how to deal with the potential for double counting. We would like for you to explain
your understanding of this issue, considering the following bulleted questions, and then
answer a few quantitative questions.


How does this potential arise?
Which papers influence your views about this question?
Now please answer the following questions:
Question
Setting
7
US
Exposure
(Effect Interval)
Interaction (1 week)
Change
Pollutant
Composition
Baseline
1 μg/m3
PM2.5
Ambient
18 μg/m3
What fraction of the 50th percentile answer you gave in Question 1 (US, long-term) was due to effects which
would express themselves within 1 week of the change in exposure? To express the uncertainty associated
with this question, please provide the 5th, 25th, 50th, 75th, and 95th percentiles of your estimate.
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
Version 5-3, Printed: June 16, 2004
Elicitation Protocol
8
US
Page 11 of 21
Interaction (3 months)
1 μg/m3
PM2.5
Ambient
18 μg/m3
What fraction of the 50th percentile answer you gave in Question 1 (US, long-term) was due to effects which
would express themselves within three months of the change in exposure? To express the uncertainty
associated with this question, please provide the 5th, 25th, 50th, 75th, and 95th percentiles of your estimate.
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
Version 5-3, Printed: June 16, 2004
Elicitation Protocol
Page 12 of 21
Differential Toxicity
Several authors have discussed the notion that different constituents of PM2.5 may have
different toxic effects and magnitudes. This possibility is of significant importance to
regulatory decision-makers because the control of different sources may remove different
constituents of PM2.5 and lead to an overall change in the composition of ambient PM2.5.
In this section, we would like to discuss your interpretation of the evidence for
differential toxicity. We are particularly interested in what you believe can and cannot
be said about quantitative estimates for this effect. We would like you to consider the
bulleted questions and answer the question that follows them. Note that if you and other
experts do not believe this is possible, all PM constituents will implicitly be treated as
equivalently toxic.



What kind of evidence would be ideal in this issue?
What evidence do we actually have?
How does this compare with the ideal evidence?
What do you believe to be the important constituents of PM2.5 as regards mortality?
For long-term exposures and ambient conditions in the US as defined in Question 1, how
would your estimates change if the reduction in PM2.5 were solely composed of
(separately) the following individual constituents?
Size of Effect
Lower
Unchanged
Primary (EC/OC)
Ammonium Sulfate
Ammonium Nitrate
Crustal Fine PM
Secondary Organics
Diesel Particulate
Burning Crude Oil
Transition Metals
Other (Specify)
Version 5-3, Printed: June 16, 2004
Higher
Less
Uncertainty
Unchanged
More
Elicitation Protocol
Page 13 of 21
If you indicated that any of these composition categories were more toxic than the
ambient mixture, then for the most toxic category, how would you revise your estimate of
uncertainty for the long-term exposure conditions defined below?
Question
9
Setting
US
Exposure
(Effect Interval)
Long-term
Change
1 μg/m3
Pollutant
PM2.5
Composition
[Indicate]
Baseline
18 μg/m3
What is your estimate of the true, but unknown, percent change in the total annual, non-accidental mortality
rate in the adult U.S. population resulting from a permanent 1 μg/m3 reduction of [most toxic] in long-term
annual average PM2.5 (from a population-weighted baseline concentration of 18 μg/m3) throughout the U.S.?
To express the uncertainty associated with the concentration-response relationship, please provide the 5th, 25th,
50th, 75th, and 95th percentiles of your estimate.
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
If you indicated that any of these composition categories were less toxic than the ambient
mixture, then for the least toxic category, how would you revise your estimate of
uncertainty for the long-term exposure conditions defined below?
Question
10
Setting
US
Exposure
(Effect Interval)
Long-term
Change
1 μg/m3
Pollutant
PM2.5
Composition
[Indicate]
Baseline
18 μg/m3
What is your estimate of the true, but unknown, percent change in the total annual, non-accidental mortality
rate in the adult U.S. population resulting from a permanent 1 μg/m3 reduction of [least toxic] in long-term
annual average PM2.5 (from a population-weighted baseline concentration of 18 μg/m3) throughout the U.S.?
To express the uncertainty associated with the concentration-response relationship, please provide the 5th, 25th,
50th, 75th, and 95th percentiles of your estimate.
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
Version 5-3, Printed: June 16, 2004
Elicitation Protocol
Page 14 of 21
For the following questions, please refer to the following plots and map.
Question
Setting
Exposure
(Effect Interval)
Change
Pollutant
11
Kuwait
9 months
(plot)
PM2.5
Composition
Burning
Crude
Year
1991
What is your estimate of the true, but unknown, percent change in the non-accidental mortality rate in the
Kuwaiti national population alive at the beginning of the 1991 oil fires resulting from the nine-month excursion
in PM2.5 concentration patterns as given in the following plots? Note that the source of this excursion is from
the combustion of crude oil. To express the uncertainty associated with the concentration-response relationship,
please provide the 5th, 25th, 50th, 75th, and 95th percentiles of your estimate.
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
Question
Setting
Exposure
(Effect Interval)
Change
Pollutant
12
Kuwait
9 months
(plot)
PM2.5
Composition
Burning
Crude
Year
1991
What is your estimate of the true, but unknown, number of premature deaths in the Kuwaiti national
population alive at the beginning of the 1991 oil fires resulting from the nine-month excursion in PM2.5
concentration patterns as given in the following plots? Note that the source of this excursion is from the
combustion of crude oil. To express the uncertainty associated with the concentration-response relationship,
please provide the 5th, 25th, 50th, 75th, and 95th percentiles of your estimate.
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
Additional task about question 12.
We ask you to do the following: Make a calculation of a plausible “upper estimate” of the
answer to question 12 and write your approach, reasoning and calculations down. Then,
do a similar thing to calculate a plausible “lower estimate”, and a “central estimate.” We
ask you to formalize this in a 2-3 page write up. We hope that you tell how you would
weave together the answers from all the previous questions to approach this question. The
"upper, lower, and central" correspond to 95th, 5th, and 50th percentiles, respectively, but
are not necessarily equal to them. You may change your answers to question 12 after this
additional task, if you think it is warranted.
Version 5-3, Printed: June 16, 2004
Elicitation Protocol
Page 15 of 21
Concentration Time-Series at 3 Populated Areas in Kuwait
Time Series of PM10 at Jahra
1000
Ambient (Measured)
900
PM10 concentration (ug/m3)
Ambient Average = 284
800
Contribution from Oil Fires
(Modeled)
700
Oil Fire Average = 24.2
600
500
400
300
284
200
100
0
2/1/1991
24.2
3/23/1991
5/12/1991
7/1/1991
8/20/1991
10/9/1991
11/28/1991
Date
Time Series of PM10 at Mansouria
1000
Ambient (Measured)
900
Ambient Average = 300
800
PM10 concentration (ug/m3)
Contribution from Oil Fires
(Modeled)
700
Oil Fire Average = 17.3
600
500
400
300
300
200
100
0
2/1/1991
17.3
3/23/1991
5/12/1991
7/1/1991
8/20/1991
Date
Version 5-3, Printed: June 16, 2004
10/9/1991
11/28/1991
Elicitation Protocol
Page 16 of 21
Time Series of PM10 at Riqa
1000
Ambient (Measured)
900
Ambient Average = 251
PM10 concentration (ug/m3)
800
700
Contribution from Oil
Fires (Modeled)
Oil Fire Average = 28.2
600
500
400
300
251
200
100
0
2/1/1991
28.2
3/23/1991
5/12/1991
7/1/1991
8/20/1991
Date
Version 5-3, Printed: June 16, 2004
10/9/1991
11/28/1991
Elicitation Protocol
Page 17 of 21
Map of Locations Corresponding to 3 Sites
Jahra
Mansouria
Riqa
Version 5-3, Printed: June 16, 2004
Elicitation Protocol
B.
Page 18 of 21
Seed/Calibration Variables
The purpose of the following questions is to address empirical control on the quantitative
uncertainty assessments for the variables of interest. They may be used to validate the
expert assessments and to construct a set of weights for combination of these
assessments. Each question below has an answer which either has been or soon will be
observed and reported. You are asked to provide your best estimates for each answer in
the form of percentiles representing a distribution of uncertainty.
The following seed questions are based on PM10 data from London and Athens for the
period 1997 – 2001.
LONDON STATIONS
HILLINGTON
ELTHAM
BRENT
BLOOMSBURY
BEXLEY
MARUSSI
THRAKOMAKEDONES
ZOGRAFOU
ATHENS STATIONS
AGIA
PARASKEVI
LYKOVRISI
Questions on PM10 Reference station exceedances
1. On how many days in 2001 did the daily average PM10 concentration exceed 50μg/m3 at least one of the
above London stations (max 365)?
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
2. On how many days in 2001did the daily average PM10 concentration fall below 30 μg/m3 at all of the above
London stations (max 365)?
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
Version 5-3, Printed: June 16, 2004
Elicitation Protocol
Page 19 of 21
3. On how many days in 1997 did the daily average PM10 concentration exceed concentration exceed 50μg/m3
at least one of the above London stations (max 365)?
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
4. On how many days in 1997did the daily average PM10 concentration fall below 30 μg/m3 at all of the above
London stations (max 365)?
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
5. On how many days in 2001 did the daily average PM10 concentration exceed 50μg/m3 at least one of the
above Athens stations (max 365)?
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
6. . On how many days in 2001did the daily average PM10 concentration fall below 30 μg/m3 at all of the above
Athens stations (max 365)?
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
Questions on Mortality
LONDON
The annual average for the year 2000 of daily PM10 concentration, averaged over the 5
stations in London was 18.4 g/m3. The highest weekly concentration averaged over the
London stations was 33.4 g/m3
Version 5-3, Printed: June 16, 2004
Elicitation Protocol
Page 20 of 21
7. What is the ratio:
Number of non-accidental deaths in the week (7 days starting from January 1st) of 2000 with the highest
average PM10 concentration
/
Weekly average number non-accidental deaths in 2000.
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
8. What is the ratio
Number of cardiovascular deaths (ICD10 Cause I) in the week (7 days starting from January 1st) of 2000 with
the highest average PM10 concentration
/
Weekly average number cardiovascular deaths in 2000
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
ATHENS
The annual average for the year 2001 of daily PM10 concentration, averaged over the 5
Athens stations was 45.5 g/m3; the highest weekly concentration averaged over the
Athens stations was 63.2 g/m3; the lowest weekly concentration averaged over the
Athens stations was 16.5 g/m3.
9. What is the ratio
Number of non-accidental deaths in the week (7 days starting from January 1st) of 2001 with the highest
average PM10 concentration
/
Weekly average number non-accidental deaths in 2001
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
10 What is the ratio:
Number of cardiovascular deaths in the week (7 days starting from January 1st) of 2001 with the highest
average PM10 concentration
/
Weekly average number cardiovascular deaths in 2001
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
Version 5-3, Printed: June 16, 2004
Elicitation Protocol
Page 21 of 21
11. What is the ratio
Number of non-accidental deaths in the week (7 days starting from January 1st) of 2001 with the lowest average
PM10 concentration
/
Weekly average number non-accidental deaths in 2001
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
12 What is the ratio
Number of cardiovascular deaths in the week (7 days starting from January 1st) of 2001 with the lowest average
PM10 concentration
/
Weekly average number cardiovascular deaths in 2001
5% :____________ 25%:____________ 50% :____________ 75%:____________ 95%:____________
Version 5-3, Printed: June 16, 2004