Why must the ordering variable in a Goldfeld–Quandt test be chosen thoughtfully?
StatisticsLevel: IntermediateRegression diagnostics
Why must the ordering variable in a Goldfeld–Quandt test be chosen thoughtfully?
Hints
- The alternative is order-dependent.
- Avoid post-hoc search.
Answer
The test compares residual variances in ordered subsets. An irrelevant order can hide the hypothesized variance change; choosing it after many trials inflates false discoveries.
Source and licence: statsmodels — model diagnostics · BSD-3-Clause · answer: Quantoros editorial