Functions to calculate contributions to the score vector from individual observations for a fitted model object.
# S3 method for Bernoulli
estfun(x, ...)
# S3 method for GP
estfun(x, eps = 1e-05, m = 3, ...)A fitted model object.
Further arguments. None are used for
estfun.Bernoulli or estfun.GP.
These control the estimation of the observed
information in gpObsInfo when the GP shape parameter \(\xi\) is
very close to zero. In these cases, direct calculation is unreliable.
eps is a (small, positive) numeric scalar. If the absolute value
of the input value of \(\xi\), that is, pars[2], is smaller than
eps then we approximate the [2, 2] element using a Taylor
series expansion in \(\xi\), evaluated up to and including the
mth term.
An \(n \times k\) matrix containing contributions to the score function from \(n\) observations for each of the \(k\)
parameters.
estfun.Bernoulli: an \(n \times 2\) matrix, where
\(n\) is the sample size, the length of the input data to
fitBernoulli. The column is named prob.
estfun.GP: an \(n \times 2\) matrix, where \(n\) is the
sample size, the length of the input data to fitGP.
The columns are named sigma[u] and xi.
An estfun method is used by
meatCL to calculate the
meat in the sandwich covariance estimator on which
the log-likelihood adjustments in flite are based.
Specifically, meatCL is used to calculate
the argument V passed to adjust_loglik.
Bernoulli for maximum likelihood inference for the
Bernoulli distribution.
generalisedPareto for maximum likelihood inference
for the generalised Pareto distribution.
library(sandwich)
# estfun.Bernoulli
bfit <- fitBernoulli(c(exdex::cheeseboro) > 45)
head(estfun(bfit))
#> prob
#> [1,] -1.0285
#> [2,] -1.0285
#> [3,] -1.0285
#> [4,] -1.0285
#> [5,] -1.0285
#> [6,] -1.0285
# estfun.generalisedPareto
gpfit <- fitGP(c(exdex::cheeseboro), u = 45)
head(estfun(gpfit))
#> sigma[u] xi
#> [1,] -0.063902582 -0.3512112
#> [2,] -0.086318699 -0.1962096
#> [3,] -0.075227392 -0.2790858
#> [4,] -0.063902582 -0.3512112
#> [5,] -0.003503566 -0.5330688
#> [6,] -0.075227392 -0.2790858