confint() is applicable to an EMLT
object to produce simultaneous confidence intervals.
All model objects gain control slot of
ControlEL class. All methods that apply to these objects
inherit control unless it is overwritten by the user
explicitly.
summary() is applicable to an object that inherits
from EL, ELT, and EMLT.
A more informative message is printed regarding the convergence status.
optim slot in all model or summary objects gains a
single logical element cstr that shows whether a
constrained EL computation is involved or not.
logLik() is deprecated and will be removed in a future
release.confreg() checks whether parm matches the
parameters in object when a character vector
is specified for parm.New accessor method logProb() extracts a model’s log
probabilities of empirical likelihood.
el_lm() and el_glm() gain an argument
offset.
el_glm() accepts quasipoisson family
with "identity" link function for the argument
family.
elt() accepts a character vector for the argument
lhs, allowing a symbolic description of a
hypothesis.
eltmt() accepts a character vector as an element of
the argument lhs, allowing a symbolic description of
hypotheses.
plot() applies to an object that inherits from
EL to plot empirical likelihood displacement values versus
observation index.
New dataset thiamethoxam added.
coef() and getDF() is applicable to an
object of class EMLT.
print() shows the tested hypothesis when applied to
an object of class ELT.
print() shows the tested hypotheses, the estimates,
and marginal degrees of freedom when applied to an object of class
ELMT. The description of the hypotheses and the estimates
are printed only when the marginal degrees of freedom are all
one.
"boot" option in the calibrate argument
of elt() yields a more reliable result when applied to an
object that inherits from LM.
Internal routines for projection operation do not compute an explicit inverse (thanks to @awstringer1).
elmt() returns a correct critical value when applied
to an object of class QGLM.
"boot" option in the calibrate argument
of elt() works with an object of class
SD.
el_glm() accepts quasipoisson family
with "log" link function for the argument
family.
New accessor methods added (chisq(),
critVal(), getDF(), getOptim(),
sigTests(), logL(), and
pVal()).
conv() is applicable to an object returned by
summary().
print() shows class-specific information.
p.value returned by el_eval() is
renamed to pval for consistency with other
functions.
confint() and confreg() are not applicable
to an object whose data is NULL.el_mean() takes arguments in a different order to
comply with the ‘tidyverse’ style. It takes the data argument
x first, followed by the parameter specification
par as el_mean(x, par).
lht() is renamed to elt().
model argument in el_mean(),
el_lm(), and el_glm() are removed. Use
keep_data in el_control().
New package dependencies are added (BH, dqrng, and graphics).
New elt() replaces lht(). It accepts
additional arguments alpha and
calibrate.
New el_sd() performs empirical likelihood test for
the standard deviation.
New elmt() tests multiple hypotheses with empirical
likelihood.
New weights() extracts the re-scaled weights from a
model.
New formula() extracts the model formula used from a
model.
New nobs() extracts the number of observations from
a model.
New conv() extracts the convergence status from a
model.
New logLR() extracts the log empirical likelihood
ratio from a model.
el_control() gains additional arguments
verbose, keep_data, seed,
b, and m.
cv argument in confint() and
confreg() defaults to NULL. If
non-NULL, level is ignored.
probit link produces a more accurate result in
el_glm().
print() for an EL object shows whether
the data are weighted or not.
All row or column names (if any) of input data are preserved in a
fitted EL object.
confint() and confreg() check if the
cv argument is compatible with the th value
set by control_el().lht() accepts both numeric vector and matrix for
lhs and rhs arguments.
OpenMP parallelization is available for confint() by
specifying nthreads through control
argument.
el_test() is removed.
el_pairwise() is deprecated and will be removed in a
future release.
S4 classes, generics, and methods are adopted throughout the package.
New confreg() constructs confidence
regions.
New eld() computes empirical likelihood displacement
values.
New el_control() the specifies control
argument.
New el_glm() performs empirical likelihood tests to
generalized linear models. More families and link functions will be
supported in a future release.
confint() gains cv argument for a
user-supplied critical value.
el_aov() is removed.
el_test() is deprecated and will be removed in a
future release.
New lht() performs linear hypothesis
testing.
New confint() constructs confidence
intervals.
New logLik() extracts empirical
log-likelihood.
el_aov() is deprecated in favor of
el_lm(). It will be removed in a future release.el_eval() is added for direct computation with
custom estimating functions.
el_mean() and el_lm() accepts an
optional weights argument for weighted EL. Arguments on
optimization are now handled by a new control argument. It
will be used in other functions in future releases.
el_lm() performs empirical likelihood tests for
linear models.el_aov() performs a one-way analysis of
variance.