When SciPy estimates Box–Cox lambda and alpha=0.05 is supplied, what does the extra returned interval describe?
StatisticsLevel: IntermediateDistribution checking
When SciPy estimates Box–Cox lambda and alpha=0.05 is supplied, what does the extra returned interval describe?
Hints
- Separate uncertainty in a parameter from spread of observations.
- Coverage is 1-alpha; identify which parameter is optimized.
Answer
It is a 95% likelihood-based confidence interval for the transformation parameter lambda, not a prediction interval for transformed observations. A wide interval indicates weak information about the transformation strength under the model assumptions.
Source and licence: SciPy — distribution diagnostics and robust tests · BSD-3-Clause · answer: Quantoros editorial