xxx

1 2 | ```
effectiveCurvature(cur, QRofD, residual, s.sqr, show.details=FALSE,
warn=TRUE)
``` |

`cur` |
the relative curvature array. See |

`QRofD` |
QR decomposition of D array from Bates and Watts. |

`residual` |
point in sample space where the curvature should be calculated (possibly shifted by subtracting actual data to give a residual). |

`s.sqr` |
sample estimate of the residual variance. |

`show.details` |
logical indicating if intermediate calculations should be printed. |

`warn` |
see details. |

Effective residual curvature from Bates and Watts p260 Calculate the scaled RMS curvatures relative to a confidence disk radius and extreme axis ratios. ref. Bates and Watts (1983) p254 eqn (7.23). and Bates and Watts J.R. Statist.Soc. B (1980).

Transform the residual vector by multiply Q-transpose and sqrt(s.sqr*p). Calculate the p by p effective residual curvature matrix B and its eigenvalues. Bates and Watts p260

... If I-B is not positive definite, where B is the effective residual curvature matrix, and warn is TRUE, then a warning will indicate that the calculation does not seem to correspond to a local minimum.

x

Bates and Watts (1980), J.R. Statist.Soc. B. Bates and Watts( 1983), 'Nonlinear Regression Analysis and Its Applications.'

`curvature`

`project`

`curvatureStats`

`relCurvature`

Questions? Problems? Suggestions? Tweet to @rdrrHQ or email at ian@mutexlabs.com.

All documentation is copyright its authors; we didn't write any of that.