Hybrid System Science Methods: Some Observations Nathaniel Osgood - - PowerPoint PPT Presentation

SLIDE 1

Hybrid System Science Methods: Some Observations

Nathaniel Osgood Agent-Based Modeling Bootcamp for Health Researchers August 25, 2011

SLIDE 2

System Science Methodologies: Highly Complementary

• Different modeling methodologies seek to answer

different types of questions

• No one system science methodology offers a

replacement for the others

• Significant synergies can be secured by using

combinations of methodologies to address the same problem

– As cross-checks on understanding where two or more can be applied – Exploiting competitive advantages

SLIDE 3

Multi-Framework Modeling

• We have found the use of multiple frameworks

highly effective

– Co-evolving multiple models for

• Cross-validation
• Asking different sorts of questions
• Revealing new questions to answer

– Within a single model

• Dealing with questions at different scales
• Improving robustness of models
• Allowing for representation & changing of factors that are
• therwise ignored

SLIDE 4 High later gr

• w

th PL Consequence1 Low later growth 1-PL Consequence2 Expand capacity High later gr

• w

th PL Consequence3 Low later growth 1-PL Consequence4 No expansion High ear ly growth PE High later gr

• w

th PL Consequence5 Low later growth 1-PL Consequence6 Expand capacity High later gr

• w

th PL Consequence7 Low later growth 1-PL Consequence8 No expansion Low early growth 1-PE Expand capacity High later gr

• w

th PL Consequence9 Low later growth 1-PL Consequence10 Expand capacity High later gr

• w

th PL Consequence11 Low later growth 1-PL Consequence12 No expansion High ear ly growth PE High later gr

• w

th PL Consequence13 Low later growth 1-PL Consequence14 Expand capacity High later gr

• w

th PL Consequence15 Low later growth Consequence16 No Expansion Low early growth 1-PE No expansion Initial Decision

Discrete Event Modeling Agent- Based Modeling Social Network Analysis Decision Analysis

Reminder:Multiple Model Types

System Dynamics

Normal and Underweight Weight Overweight Pregnant with GDM History of GDM T2DM Developing Obesity Pregnant Normal Weight Mothers with No GDM History Completion of Pregnancy to Non-Overweight State Completion of GDM Pregnancy Women with History of GDM Developing T2DM Overweight Individuals Developing T2DM Normal Weight Individuals Developing T2DM Pregnant with T2DM New Pregnancies from Mother with T2DM Completion of Pregnancy for Mother with T2DM Pregnant Overweight Mothers with No GDM History Pregnancies of Overweight Women Completion of Pregnancy to Overweight State Pregnancies of Non-Overweight Women Pregnancies to Overweight Mother Developing GDM Pregnancies to Non-Overweight Mother Developing GDM Pregnant with Pre-Existing History of GDM Pregnancies for Women with GDM Pregnancies Developing GDM from Mother with GDM History Completion of Non-GDM Pregnancy for Woman with History of GDM Shedding Obesity Pregnant Women Developing Persistent Overweight/Obesity Oveweight Babies Born from T2DM Mothers Pregnant Women with GDM that Continue on to Postpartum T2DM bies Born from

• ther with

GDM erweight Babies Born to regnant Normal Weight Mothers Pregnancy Duration <Birth Rate> Normal Weight Deaths Overweight Deaths T2DM Deaths Deaths from Non-T2DM Women with History of GDM

SLIDE 5

Discrete Event Modeling Agent- Based Modeling Social Network Analysis Decision Analysis

Multiple Model Types

System Dynamics

SLIDE 6

SNA Can Facilitate ABM

• Social network statistics that be used to

formulate synthetic networks

• Identify patterns for calibration &

investigation

• Cross-checks on ABM simulation findings
• Network visualization
• Highlighting diverse settings for contact

SLIDE 7

Challenges in Using Data from SNA in ABM

• SNA can provide an extremely valuable source of

data to use for grounding ABM network structure

• It is relatively easy to get networks from software

like Pajek into software like AnyLogic or Repast

• The bigger issue here is that we need to represent

the hypothesized “true” spread of infection over the network

– To do this, we need to represent the hypothesized underlying network that lies behind – Even the best of SNA data is highly incomplete (e.g. due to asymmetries in case-contact data, sampling in snowball sampling)

SLIDE 8

SNA Providing Context For ABM

SLIDE 9

Example Network Structure

SLIDE 10

Multiple Modeling Types

System Dynamics Agent- Based Modeling Social Network Analysis

SLIDE 11

Agent-Based Modeling Facilitates SNA

• Exploring dynamic hypotheses to explain SNA patterns
• Formulating ideas for SNA metrics that could be

highly effective (discriminatory) for identifying at-risk individuals

• Understanding dynamic implications of given network

structure

• Understanding implications of changing network structure
• Evaluating SNA-informed interventions (e.g. SNA-metric

prioritized contact tracing)

• Examining impact of additional collection of SNA data (e.g.

, more complete contact tracing)

• Positing possible pieces of missing structure in SNA

network

SLIDE 12

ABM To Explain Emergent Patterns Uncovered via SNA

A. Al-Azem, Social Network Analysis in Tuberculosis B. Control Among the Aboriginal Population of Manitoba2006

SLIDE 13

Multiple Modeling Types

System Dynamics Agent- Based Modeling Social Network Analysis

SLIDE 14

System Dynamics Supporting ABM

• Description of continuous individual-level

evolution

• Deriving calibrated parameter estimates for low-

level model

• Focusing AB exploration
• Qualitative diagramming of

– Interactions at a particular scale – Hypothesized drivers underlying emergent behaviour

SLIDE 15

Multiple Modeling Types

System Dynamics Agent- Based Modeling Social Network Analysis

SLIDE 16

Agent-Based Modeling in Support of SD

• Cross-validating SD aggregation: Evaluating

importance of

– Stratification by heterogeneities – Stochastics – Network dynamics

• Giving insight into feedbacks to depict
• Investigating specialized interventions

– e.g. Interventions that depend on individual history, network position, etc.

• Use to determine parameters for SD model

SLIDE 17

Multi-Scale ABM-SD Hybrid Strategies

• Agent Based & System Dynamics

– System Dynamics within ABM: Agent behaviour described w/stocks & flows (optionally, within SD tools) – ABM within System Dynamics: Agents drive some flows – Using Qualitative Methods of System Dynamics (Group Model Building, Causal Loop diagram) to elicit understanding for an agent-based model

• DES & ABM

– Agents associated with Entities – Entity presentation dependent on Agent State – Agent evolution dependent on entity Treatment

SLIDE 18

System Dynamics & Individual-Based Modeling

• Individual-based models can be created using

– Traditional System Dynamics software

• Small populations:

– Separate stocks for each individual – Hand-drawn connections

• Larger Populations

– Subscripting stocks by population member – Binary network matrices

– Stock & flows in other dynamic modeling software

• e.g. in AnyLogic

– System Dynamics methodology

• Feedback-centric reasoning
• Process-based work

SLIDE 19

Network Embedded Individuals

Uninfected Cells Infected Cells Virus Load Uninfected Cell Replentishment New Cell Infections Uninfected Cell death Infected Cell Death Virion Production From Infected Cells Virion Clearance Uninfected Cell Replentishment Rate Mean Infected Cell Lifetime Mean Uninfected Cell Lifetime Mean Virion Lifetime Likelihood Density of Infection by Single Virion Per Infected CellVirion Production Rate Virion Production Rate Per Contact Virions Rate 1 Person Mean Viral Load <Population Size> Mean Uninfected Cells Mean Infected Cells <Population Size> <Population Size> Mean of Viral Load

• f Neighbors

CTLs immune response to infected cells CTL turnover CTL responsiveness Mean CTL lifespan infected cell death by CTLs rate which infected cells are killed by CTLs Virion Production Rate if Non Quantized Infection

SLIDE 20

Individual-Based Model in Vensim

All of these stocks & their associated flo Population member (via population-mem

SLIDE 21

Population-Member Subscripting

SLIDE 22

Example Interactions between Global & Local Levels

A Global Level (Aggregate, Cross Population) Factor!

SLIDE 23

Example Individual-Level Risk Factors

An Individual-Level Risk Factor Another Individual-Level Risk Factor (here, represented categorically, but we could Represent it as a continuous variable – e.g. cumulative smoke exposure, some estimate of cumulative physiologic damage from smoke, a moving average of smoke exposure, etc.)

SLIDE 24

Impact of Risk Factors on Individual Dynamics

SLIDE 25

Multiple Modeling Types

System Dynamics Agent- Based Modeling Social Network Analysis

SLIDE 26

System Dynamics in Support of SNA

• Examining (aggregate) impact of SNA-driven

feedbacks

• Understanding dynamic implications of certain

levels of contact tracing

• Dynamics within an individual
• Coupled individual dynamics
• Identifying key parameters for SNA to examine

– E.g. Recognizing importance of a small amount of contact tracing

SLIDE 27

Discrete Event Modeling Agent- Based Modeling Social Network Analysis Decision Analysis

Multiple Model Types

System Dynamics

SLIDE 28

Agent-Based Modeling in Support of DES

• Representing network of individuals in population
• utside of flow process

– Prior to entry (development of conditions) – Following exit (e..g trajectory dependent on quality of care) – Routing inflowing agents process based on agent’s history of care – e.g. representing “catchment basin” of care facility

SLIDE 29

ABM & DES

SLIDE 30

Discrete Event Modeling Agent- Based Modeling Social Network Analysis Decision Analysis

Multiple Model Types

System Dynamics

SLIDE 31

31

Two Relevant Methodologies

• Decision Analysis

– Good for decision problems under uncertainty w/known scenario consequences

• No endogenous means of

determining consequences

– Characterizes structured policy space – Sophisticated statistical tools & Sensitivity analyses typical – Identification of robust strategies via backwards induction – Discrete

• Events/decisions
• Time
• Dynamic Modeling

– Good for representing complex system response to scenario (events and actions) – Policy representation

• Highly flexible
• Less structured policy space

– Basic statistical tools – Potentially continuous

• Time
• Events/decisions

SLIDE 32

32

Decision Tree To Structure Policy Space

High later gr

• w

th

PL

Consequence1 Low later growth

1-PL

Consequence2 Expand capacity High later gr

• w

th

PL

Consequence3 Low later growth

1-PL

Consequence4 No expansion High ear ly growth

PE

High later gr

• w

th

PL

Consequence5 Low later growth

1-PL

Consequence6 Expand capacity High later gr

• w

th

PL

Consequence7 Low later growth

1-PL

Consequence8 No expansion Low early growth

1-PE

Expand capacity High later gr

• w

th

PL

Consequence9 Low later growth

1-PL

Consequence10 Expand capacity High later gr

• w

th

PL

Consequence11 Low later growth

1-PL

Consequence12 No expansion High ear ly growth

PE

High later gr

• w

th

PL

Consequence13 Low later growth

1-PL

Consequence14 Expand capacity High later gr

• w

th

PL

Consequence15 Low later growth

1-PL

Consequence16 No Expansion Low early growth

1-PE

No expansion Initial Decision

Time Decision node Event node Terminal Node (Consequence for Scenario) Scenario

SLIDE 33

33

Computed By System Dynamics Model

Backwards Induction

• Run SD model to

determine set of (preference- adjusted) outcomes for all terminal nodes

• Compute expected

value & risk profiles at event nodes

• At decision nodes,

choose decision with highest value

• FAST

(PL*Consequence1+(1-PL)*Consequence2) (PL*Consequence3+(1-PL)*Consequence4) High later gr

• w

th

PL

Consequence1 Low later growth

1-PL

Consequence2 Expand capacity High later gr

• w

th

PL

Consequence3 Low later growth

1-PL

Consequence4 No expansion High ear ly growth

PE

High later gr

• w

th

PL

Consequence5 Low later growth

1-PL

Consequence6 Expand capacity High later gr

• w

th

PL

Consequence7 Low later growth

1-PL

Consequence8 No expansion Low early growth

1-PE

Expand capacity High later gr

• w

th

PL

Consequence9 Low later growth

1-PL

Consequence10 Expand capacity High later gr

• w

th

PL

Consequence11 Low later growth

1-PL

Consequence12 No expansion High ear ly growth

PE

High later gr

• w

th

PL

Consequence13 Low later growth

1-PL

Consequence14 Expand capacity High later gr

• w

th

PL

Consequence15 Low later growth

1-PL

Consequence16 No Expansion Low early growth

1-PE

No expansion Initial Decision

Backwards Induction

SLIDE 34

34

Sensitivity Analysis

• Offline sensitivity of

likelihood impact on strategy selection

• Offline analysis

(including Monte Carlo)

• f impact of likelihood

change on risk profiles

SLIDE 35

Population Subscripting Tradeoffs

Advantages

• Conceptually simple
• Can SD tools

– State trajectory file recording – Easy construction, structure visualization

• No programming

– Sensitivity analysis – Easy to aggregate

Disadvantages

• Difficult to visualize network

structure & spread or spatial embedding

• Awkward to realize changing

population size