attachVarPrep: attachVarPrep
In twNlme: Standard errors for basic nlme and gnls model predictions

Description Usage Arguments Details Value Author(s) See Also Examples

Attach partial derivative and residual variance functions to the nonlinear fitted object

attachVarPrep(object, form = formula(object), fDerivFixef = NULL, 


    fDerivRanef = NULL, fVarResidual = NULL, fAddDummies = NULL)

`object`	the fitted nlme object
`form`	the formula used to fit the object, either formula or string used for automated derivation
`fDerivFixef`	`function(nfit,newdata,pred)` of derivatives in respect to fixed effects at newdata
`fDerivRanef`	`function(nfit,newdata,pred)` of derivatives in respect to random effects at newdata
`fVarResidual`	`function(nfit,newdata,pred)` to calculate var(residual) at newdata
`fAddDummies`	`function(newdata)` see "Handling categorial variables"

For usage with varPredictNlmeGnls, this function attaches to the fitted object

derivative functions
residual variance function
Variance-Covariance methods

Automatic derivation

If proper basic formula is given, fDerivFixef and fDerivRanef will be automatically derived from the model.

Up until now, only single level random effects models are supported for automatic derivation.

Handling of categorial variables

item fAddDummies=function(newdata) of restult entry varPrep adds columns for dummy variables

of categorial variables to newdata.

Default implementation supports only (and assumes) contr.treatment coding.

For other codings user must provide the function with argument fAddDummies.

Variance of Residuals

Providing no argument fResidual assumes iid residuals, i.e. weights=NULL.

For other residual variance models. See e.g. varResidPower corresponding to weights=varPower(form=~fitted(.))

nfit with additional entry varPrep, which is a list of

`varFix`	variance-covariance matrix of fixed effects
`varRan`	variance-covariance matrix of random effects
`coefFix`	names of the fixed coefficients in gradiant function
`coefRan`	names of the random coefficients in gradiant function
`gradFix`	derivative function for fixed effects
`gradRan`	derivative function for random effects
`fAddDummies`	function to add dummy columns for categorial variables to predictor data frame
`fVarResidual`	function to calculate residual variance
`fullFormula`	the extended formula as a string

Thomas Wutzler

twNlme-package

data(modExampleStem)


#mtrace(.covarMap)


nfit <- attachVarPrep( modExampleStem, form = "b0*dbh^b1*height^b2")





data(Wutzler08BeechStem)


newdata=data.frame(dbh=18.8, height=16.9, age=40, si=30, alt=470)


(uNew <- nfit$varPrep$gradFix(newdata=newdata))


(wNew <- nfit$varPrep$gradRan(newdata=newdata))





newdata=head(Wutzler08BeechStem)


(uNew <- nfit$varPrep$gradFix(newdata=newdata))


(wNew <- nfit$varPrep$gradRan(newdata=newdata))