Transcript A Discussion: Random Thoughts and Risky Propositions

A Discussion:
Random Thoughts and
Risky Propositions
Sheldon H. Jacobson
Director, Simulation and Optimization Laboratory
Department of Computer Science
University of Illinois
Urbana, IL
[email protected]
https://netfiles.uiuc.edu/shj/www/shj.html
SAMSI 9/16-19/2007
Workshop (RTP, NC)
(C) Jacobson 2007
1
Lianne Sheppard
Environmental Health Modeling
•Methods to measure and identify
environmental effect / risk on health.
•Important Problem
–Large number and amount of substances that
can be scrutinized.
–Important policy and economic implications.
SAMSI 9/16-19/2007
Workshop (RTP, NC)
(C) Jacobson 2007
2
Anne Smith
Environmental Risk Assessment
•Risk Assessment for Ambient Air Pollutants
•Important Problem
–Air quality can be measured by a large quantity
of substances / toxins.
–Numerous sources of uncertainty in the process.
–Important policy and economic implications.
SAMSI 9/16-19/2007
Workshop (RTP, NC)
(C) Jacobson 2007
3
A Simple Schematic
Environmental
Risks
Black
Box
(Natural,
Man-made)
(Human, Animal,
Birds, Insects)
Mortality
Morbidity
Geologic
Industrial
SAMSI 9/16-19/2007
Workshop (RTP, NC)
Health
(Chronic, Acute)
(C) Jacobson 2007
4
The Analysis Process
•Models, Models, Models (Environmental Health Modeling)
–Disease
•Quantifies the true environmental exposure to the disease
outcome
– Exposure
•Captures the distribution of exposure over space, time, and
individuals
– Measurement
•Quantifies measured exposure to the true unknown exposure
•Data, Data, Data….
–Quality, quantity, cleanliness
•Not always clear what one is getting
SAMSI 9/16-19/2007
Workshop (RTP, NC)
(C) Jacobson 2007
5
Observations and
“Food for Thought”
•Model simplicity versus data complexity
–Is it better to have a complex model with little data
available or a simple model with much data available?
• Model Validation and Verification is a challenge
–Invisible (environmental, personal, policy) biases can
creep into the analysis.
–Can such biases cloud what one is trying to measure
/ identify?
–How does one separate the cause/ effect relationship
from system noise?
SAMSI 9/16-19/2007
Workshop (RTP, NC)
(C) Jacobson 2007
6
Observations and
“Food for Thought”
•Design of Experiment
–Numerous challenges.
–Input controls are not that easy to control.
•Fewer questions can lead to more insight
–Focus study on particular relationship(s).
–Are focused studies even possible?
–Breadth versus depth of analysis.
SAMSI 9/16-19/2007
Workshop (RTP, NC)
(C) Jacobson 2007
7
Observations and
“Food for Thought”
•Static versus temporal associations
–Must both be addressed?
–Knowing “when” may be as challenging as knowing “if”.
•Many questions can be posed.
–A “substance” causes what “conditions”?
–A “condition” is caused by what “substances”?
–Knowing “If” and “how much” may both be critical.
–Which questions should be addressed?
SAMSI 9/16-19/2007
Workshop (RTP, NC)
(C) Jacobson 2007
8
Observations and
“Food for Thought”
•Which error is most dangerous?
–Not identifying an effect that exists (false clear) or
believing that an effect exists which does not (false alarm)?
–Policy implications may have “long legs”.
–Complex system implications.
•The goal may change.
– Are we looking for a “needle in a haystack”, or should we
ask why needles keeps ending up in a haystack, or in a
particular section of a haystack?
SAMSI 9/16-19/2007
Workshop (RTP, NC)
(C) Jacobson 2007
9
Contemporary Issues
•Bioterrorism agent monitoring
•Pandemic influenza, infectious diseases and
emerging pathogens
–Avian flu (H5N1)
•Prevention, detection, treatment
•Disease monitoring / epidemiology
–Can we create models that serve as “canaries in a
mine shaft?”
SAMSI 9/16-19/2007
Workshop (RTP, NC)
(C) Jacobson 2007
10
Key Observation
There are many more questions than answers.
?
SAMSI 9/16-19/2007
Workshop (RTP, NC)
(C) Jacobson 2007
11