Abstract
We investigate the biases of the residuals and maximum likelihood parameter estimates from normal nonlinear regression models. Emphasis is placed on a class of partially nonlinear models, on the role of individual cases in determining bias, on how bias affects standard diagnostic methods, and on the relationship between bias and curvature.
Original language | English (US) |
---|---|
Pages (from-to) | 615-623 |
Number of pages | 9 |
Journal | Biometrika |
Volume | 73 |
Issue number | 3 |
DOIs | |
State | Published - Dec 1 1986 |
Keywords
- Diagnostic
- Influence
- Intrinsic curvature
- Residual
- Tarnsformation