Skip to content

Practice

Exercises

Each exercise is graded on the server with a published rubric. Signed-in learners keep a record of every attempt.

Causal identification12

Causal identification

Adjustment set

Loyalty affects both discount assignment and purchase. Should loyalty be adjusted for when estimating the effect of the discount?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Targeted Promotion with Virtual Buyers
Causal identification

Risk versus treatment effect

Patient X has 40% readmission risk and an estimated support effect of −2 points. Patient Y has 20% risk and an effect of −8 points. With one slot, who should receive support to avoid the most readmissions?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Hospital Flow and Personalized Discharge
Causal identification

Colliders and confounders

A chargeback flag is caused by both fraud and device sharing. What happens if you condition on the flag when studying device sharing and fraud?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Fraud Contagion and Intervention Policy
Causal identification

Prediction versus intervention

Which question requires a causal model rather than a predictive one?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Hospital Flow and Personalized Discharge
Causal identification

Confounders for incentive response

Buildings on time-of-use tariffs joined past events more often and also cut more load. Which set should be adjusted for when estimating the rebate effect?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Demand Response and Grid Resilience
Causal identification

Interference between neighborhoods

Adaptive signals in District A push traffic into District B. What does this mean for estimating the effect of signals?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Urban Traffic and EV Charging Coordination

Showing 1–6 of 12

Effect estimation4

Effect estimation

CATE calculation

Purchase probability in a segment rises from 0.20 to 0.32 under a 10% discount. What is the uplift in percentage points?

Answers are checked anonymously unless you sign in.
From showcase: Targeted Promotion with Virtual Buyers
Effect estimation

Two-period difference-in-differences

Treated regions go from 60% to 66% employment; comparison regions go from 58% to 61%. What is the DiD estimate in points?

Answers are checked anonymously unless you sign in.
From showcase: Policy Shock Event Study
Effect estimation

Mediated effect

A campaign raises adoption by 10 points in total. The natural direct effect is 6 points. What is the indirect effect through peer exposure, in points?

Answers are checked anonymously unless you sign in.
From showcase: New Product Adoption through Social Networks
Effect estimation

One building CATE

An office-class building is predicted to use 820 kWh in the event window without a rebate and 742 kWh with it. What is the CATE in kWh reduction?

Answers are checked anonymously unless you sign in.
From showcase: Demand Response and Grid Resilience

Simulation design1

Simulation design

Simulation selection

Orders wait for a finite number of pickers and packers. Which engine should own this mechanism?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Targeted Promotion with Virtual Buyers

Systems thinking3

Systems thinking

Feedback reasoning

Explain why higher conversion from a promotion can reduce later customer retention.

Answers are checked anonymously unless you sign in.
From showcase: Targeted Promotion with Virtual Buyers
Systems thinking

Trace overtime feedback

Explain the feedback loop that can make overtime grow over time even when demand is stable.

Answers are checked anonymously unless you sign in.
From showcase: Workforce Scheduling, Retention and Burnout
Systems thinking

Trace fire-sale feedback

Explain how fast liquidation can make a liquidity shortfall worse.

Answers are checked anonymously unless you sign in.
From showcase: Portfolio Liquidity and Market Stress Response

Uncertainty6

Uncertainty

Monte Carlo

Why is a single simulation run insufficient for a policy decision?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Targeted Promotion with Virtual Buyers
Uncertainty

Distributions over means

Two policies have the same expected loss. Policy A has a worst-decile loss of 14M$, B has 9M$. Which is more robust?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Resilient Multitier Supply Network
Uncertainty

Interpret violation probability

A policy shows a 4% feeder overload probability across 2,000 simulated evenings. What does that mean?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Demand Response and Grid Resilience
Uncertainty

Explain correlation

Two portfolios have the same expected loss. One has losses that are spatially correlated. Explain why it needs larger reserves.

Answers are checked anonymously unless you sign in.
From showcase: Catastrophe Insurance Exposure and Claims Operations
Uncertainty

Compare correlation assumptions

Under normal correlations the fund covers obligations in 98% of paths; under the correlation-break case, 86%. What does this tell you?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Portfolio Liquidity and Market Stress Response
Uncertainty

Interpret the shortfall distribution

Why does the policy target obligation coverage in the worst decile rather than the mean?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Portfolio Liquidity and Market Stress Response

Decision optimization1

Decision optimization

Policy choice

Policy A yields more mean profit but has 45% stockout risk. Policy B yields slightly less profit and 15% risk. The limit is 20%. Which policy should you recommend?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Targeted Promotion with Virtual Buyers

Operations1

Operations

Reorder point

Weekly demand is 400 units, lead time is 3 weeks and safety stock is 500 units. What is the reorder point in units?

Answers are checked anonymously unless you sign in.
From showcase: Resilient Multitier Supply Network

Networks1

Networks

Homophily bias

Friends of adopters adopt at twice the base rate. Why is this not, by itself, evidence of peer influence?

Answers are checked anonymously unless you sign in.
From showcase: New Product Adoption through Social Networks

Digital twins3

Digital twins

Invalid effect-to-twin mappings

A segment has baseline purchase probability 0.20 and CATE +0.12. Which twin rule double-counts baseline risk?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Targeted Promotion with Virtual Buyers
Digital twins

Discount on three customer twins

Three twins have uplifts of 0.12, 0.05 and 0.01 under a discount. What are the expected incremental purchases if all three receive the offer?

Answers are checked anonymously unless you sign in.
From showcase: Targeted Promotion with Virtual Buyers
Digital twins

Update a community twin

A field report says a community’s population is 30% higher than estimated because of displaced arrivals. How should the community twin be updated?

Answers are checked anonymously unless you sign in.
From showcase: Humanitarian Disaster Response and Last-Mile Allocation

Network reasoning3

Network reasoning

Trace overload propagation

Explain how pre-cooling buildings on one feeder branch can still cause an overload later in the evening.

Answers are checked anonymously unless you sign in.
From showcase: Demand Response and Grid Resilience
Network reasoning

Mark critical dependencies

The directory server is a high-centrality node, and payroll and customer login both depend on it. Explain why isolating it is risky and what to check first.

Answers are checked anonymously unless you sign in.
From showcase: Cyberattack Propagation and Defense Prioritization
Network reasoning

Identify the route bottleneck

After the bridge closes, four communities can only be reached by a detour that doubles travel time. What is the binding constraint for those communities?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Humanitarian Disaster Response and Last-Mile Allocation

Scenario comparison11

Scenario comparison

Uniform vs targeted incentives

In the scenario lab, the uniform rebate and the combined policy both keep overload under 5%. Why is the combined policy preferred?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Demand Response and Grid Resilience
Scenario comparison

Signal timing vs pricing

The off-peak rebate alone meets the charger-wait limit but fails the accessibility limit. What is the most likely reason?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Urban Traffic and EV Charging Coordination
Scenario comparison

Compare adjuster allocations

In the lab, the adjuster surge alone meets the backlog and hardship limits. Why is it not the recommended policy?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Catastrophe Insurance Exposure and Claims Operations
Scenario comparison

Compare two schedules

The capped-overtime option cuts overtime under 25 h but fails the service level. What does the optimized schedule add to fix this?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Workforce Scheduling, Retention and Burnout
Scenario comparison

Compare dispatch rules

Full yield-priority routing gives the highest yield. Why is it not acceptable?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Semiconductor Fab Yield and Cycle Time
Scenario comparison

Compare outreach allocations

Targeted outreach alone closes the representation gap but misses the 15-month target. Why?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Adaptive Clinical Trial Recruitment and Site Operations

Showing 1–6 of 11

Calculation12

Calculation

Charger utilization

A site has 8 plugs. Over a 10-hour evening window, plugs are occupied for 56 plug-hours in total. What is the utilization in percent?

Answers are checked anonymously unless you sign in.
From showcase: Urban Traffic and EV Charging Coordination
Calculation

Modify contact rate

An effective reproduction number is 1.5. If contact rates fall by 30% and nothing else changes, what is the new value?

Answers are checked anonymously unless you sign in.
From showcase: Public Health Intervention and Epidemic Response
Calculation

Effective reproduction proxy

Transmission probability per contact is 0.05, an infected person has 10 contacts per day, and is infectious for 4 days. 40% of the population is immune. Compute R_eff.

Answers are checked anonymously unless you sign in.
From showcase: Public Health Intervention and Epidemic Response
Calculation

Expected loss for one zone

A zone has 2,000 homes insured for $300k each. Probability of damage is 0.15 and average loss given damage is 20% of insured value. What is the expected loss in $M?

Answers are checked anonymously unless you sign in.
From showcase: Catastrophe Insurance Exposure and Claims Operations
Calculation

Calculate utilization

120 contacts arrive per hour with an average handling time of 6 minutes. 14 agents are on shift. What is utilization in percent, rounded to one decimal?

Answers are checked anonymously unless you sign in.
From showcase: Workforce Scheduling, Retention and Burnout
Calculation

Compute cycle time

A fab holds 14,000 wafers in process and completes 2,000 wafers per day. Using Little’s law, what is the average cycle time in days?

Answers are checked anonymously unless you sign in.
From showcase: Semiconductor Fab Yield and Cycle Time

Showing 1–6 of 12

Diagnosis8

Diagnosis

Locate a bottleneck

Average travel time falls with adaptive signals, but charger waits do not. Explain where the bottleneck is and why signals do not fix it.

Answers are checked anonymously unless you sign in.
From showcase: Urban Traffic and EV Charging Coordination
Diagnosis

Identify the capacity threshold

Explain why the policy choice focuses on keeping peak care demand under 100% of capacity rather than only minimizing total infections.

Answers are checked anonymously unless you sign in.
From showcase: Public Health Intervention and Epidemic Response
Diagnosis

Identify the claims bottleneck

Adjuster surge clears inspections quickly, but hardship duration barely falls. Where is the bottleneck most likely to be?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Catastrophe Insurance Exposure and Claims Operations
Diagnosis

Locate a shifting bottleneck

Doubling maintenance on lithography raised yield but cycle time grew and the etch queue shrank. Explain what happened to the bottleneck.

Answers are checked anonymously unless you sign in.
From showcase: Semiconductor Fab Yield and Cycle Time
Diagnosis

Identify the site bottleneck

Site 7 has the most referrals but the lowest enrollment, and its screening wait is 34 days. What is the most likely constraint?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Adaptive Clinical Trial Recruitment and Site Operations
Diagnosis

Trace lateral movement

After the team patched the web tier, the attacker agent still reached the database. Describe how an adaptive attacker can do this and which control would cut the path.

Answers are checked anonymously unless you sign in.
From showcase: Cyberattack Propagation and Defense Prioritization

Showing 1–6 of 8

Policy decision12

Policy decision

Assess access disparity

Recommend a coordination policy for a day with one fast-charging station failure and defend the trade-off among travel time, charger access and neighborhood impact.

Answers are checked anonymously unless you sign in.
From showcase: Urban Traffic and EV Charging Coordination
Policy decision

Universal vs targeted outreach

Under the high-transmission scenario, compare universal and targeted outreach. Which would you choose, and what subgroup outcome must you report?

Answers are checked anonymously unless you sign in.
From showcase: Public Health Intervention and Epidemic Response
Policy decision

Interpret tail conditional loss

The 99% VaR is $420M and the 99% expected shortfall is $560M. Explain what each means and which should guide reserve strengthening after a severe event.

Answers are checked anonymously unless you sign in.
From showcase: Catastrophe Insurance Exposure and Claims Operations
Policy decision

Test a fairness constraint

Propose a schedule that meets the service-level and overtime limits and explain how you would check that it reduces undesirable-shift disparity without being used against individuals.

Answers are checked anonymously unless you sign in.
From showcase: Workforce Scheduling, Retention and Burnout
Policy decision

Interpret the yield and cycle-time frontier

The aggressive hybrid has higher yield than the optimized hybrid but misses the constraints. Explain the Pareto trade-off and justify the recommended choice.

Answers are checked anonymously unless you sign in.
From showcase: Semiconductor Fab Yield and Cycle Time
Policy decision

Test the representation constraint

Recommend a resource reallocation that meets the enrollment timing and representation constraints without changing protocol eligibility. Explain how representation is checked.

Answers are checked anonymously unless you sign in.
From showcase: Adaptive Clinical Trial Recruitment and Site Operations

Showing 1–6 of 12

Causal framing2

Causal framing

Identify treatment and outcome

In the trial-strip analysis, what is the treatment and what is the outcome?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Precision Irrigation and Farm Resource Planning
Causal framing

Define treatment and response

In the bench factorial experiment, which pairing correctly defines treatment and response?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Bioprocess Fermentation Yield and Scale-Up

Workflow modeling1

Workflow modeling

Deterministic vs probabilistic nodes

Which workflow component should be modeled as probabilistic rather than deterministic?

Choose one answer
Answers are checked anonymously unless you sign in.
From showcase: Agentic AI Workflow Reliability and Human Oversight