Nonstationarity Quant Interview Guide
Nonstationarity quant interview guide for changing means, variances, relationships, structural breaks, validation examples, and model caveats.
Candidates evaluating whether historical relationships stay stable.
Nonstationarity means behavior changes over time
A nonstationary process has properties that shift over time, such as mean, variance, dependence, or relationship to other variables. This weakens simple historical extrapolation.
It is common in real data
Financial and operational data can change because participants adapt, regimes shift, products change, or measurement changes. The answer should say which change matters for the model.
Concrete example
A forecast model trained during one volatility environment may produce overconfident intervals in another. The distribution of errors has changed, so historical accuracy is stale.
Respond with validation design
Use time-aware validation, regime splits, rolling estimates, monitoring, and simpler assumptions where appropriate. Nonstationarity does not make modeling impossible, but it raises the burden of evidence.
Common mistakes
Candidates often either ignore nonstationarity or declare all models useless. Strong answers identify the specific instability and propose a practical check.
Practice the pattern
Use the LeetQuidity curriculum and calibration to turn this topic into a focused practice plan.