Abstract
SUMMARY: In nonlinear calibration and regression a double point is defined with respect to estimation and to confidence sets. Loosely speaking, if a future observation is a double point then multiple point estimates or disconnected confidence regions will occur. We investigate the extent of the set of double points. A diagnostic, the probability of their occurrence, is computed.
Original language | English |
---|---|
Pages (from-to) | 33-43 |
Number of pages | 11 |
Journal | Biometrika |
Volume | 78 |
Issue number | 1 |
DOIs | |
State | Published - Mar 1991 |
Keywords
- Confidence interval
- Differential geometry
- Double point
- Estimation
- Intrinsic curvature
- Multivariate nonlinear calibration
- Nonlinear regression