Handle missing values for LGD modeling
Handling Missing Values for LGD Modeling
Context
You are building a Loss Given Default (LGD) model using account- and borrower-level features captured around the time of default. The dataset contains both continuous and categorical variables with non-trivial missingness due to reporting gaps, system migrations, and process differences across products/regions.
Task
Describe how you would handle missing values in this LGD modeling context. Specifically, compare the following approaches:
-
Multiple imputation (e.g., MICE)
-
Model-based imputation (e.g., kNN, random forest, regression)
-
Business-rule fills (domain-driven heuristics)
-
Indicator variables (missingness flags; Unknown category)
-
Leaving missingness explicit (letting the model handle NA directly)
For each, explain:
-
Assumptions about the missingness mechanism (MCAR, MAR, MNAR)
-
Pros, cons, and typical use cases in LGD modeling
-
Guardrails to avoid bias and leakage
-
How you would validate the choice and measure impact on LGD performance and stability
Constraints & Assumptions
-
Preserve the scope, facts, inputs, and requested outputs from the prompt above.
-
If the prompt leaves a detail unspecified, state a reasonable assumption before relying on it.
-
Keep the answer interview-ready: concise enough to present, but concrete enough to implement or evaluate.
Clarifying Questions to Ask
-
Clarify the task, data shape, labels, constraints, and evaluation metric.
-
State assumptions behind the math or modeling technique you choose.
-
Connect theory to practical training, debugging, and deployment implications.
What a Strong Answer Covers
-
Correct definitions and formulas where the prompt requires them.
-
A practical explanation of how the method behaves on real data.
-
Trade-offs, failure modes, diagnostics, and mitigation strategies.
-
Evaluation choices that match the product or modeling objective.
Follow-up Questions
-
How would noisy labels, class imbalance, or distribution shift affect the answer?
-
What would you monitor after deployment?
-
Which baseline would you compare against first?