forward.prediction.error {CollocInfer}

forward.prediction.error

Description

Forward prediction error objective for choice of lambda in square error criteria.

Usage

forward.prediction.error(times,data,coefs,lik,proc,pars,whichtimes=NULL)

Arguments

times

Vector observation times for the data.

data

Matrix of observed data values.

coefs

Vector giving the current estimate of the coefficients in the spline.

lik

lik object defining the observation process.

proc

proc object defining the state process.

pars

Initial values of parameters to be estimated processes.

whichtimes

Specifies the start and end times for forward prediction, given by indeces of times. This can be one of

• A list, each element of the list is itself a list of length 2; the first element gives the starting time to use and the second is a vector giving the prediction times.

• A matrix, the first column giving the starting times and the second giving the ending times.

If left NULL, whichtimes defaults to predicting one observation ahead from each observation.

Details

Forward prediction error can be used to choose values of lambda in the profiled estimation routines. The ordinary differential equation is solved starting from the starting times specified in whichtimes and measured at the corresponding measurement times. The error is then recorded. This should then be minimized by a grid search.

Value

The forwards prediction error from the estimates.

See Also

ProfileSSE, outeropt