dependences on part of the tidyverse and on
knitr are removed
maximize now has a “newton” method
lnl_ivldv now have an opposite argument
the x and y arguments of
scoretest are replaced by object and
...
ancil is replaced by instr for the coefficients of the ivldv function
commit
the vcov method now has error msg when the estimator is not available
value is now a matrix with three columns (model, saturated, null)
logLik has a sum argument (if FALSE, the individual contributions are returned)
last commit
the analytical hessian for ordreg with censored observations is fixed
a testing infrastructure is added, using the
tinytest package
for models estimated by maximum likelihood, the gradient and the
hessian are checked if check_gradient = TRUE and the result
is a check_gradient elemnt
weights are added to weibreg, ordreg,
poisreg
in ordreg, binomreg and
poisreg, the formula is replaced by terms in the
result
dmills and d2mills replaced by
mills(, deriv = 1/2)
grep and invert argument added to
select_coef
new data set random_group
two new vignettes lm_function and
micsr_objects
select_coef is now exported and has a
fixed argument (if TRUE, constant parameters
are selected), coef and vcov also have this
argument
gaze.lm.RStests replaces gaze.LMtestlist due to a naming change in the spdep package
lhs y1 | y2 replaced by y1 + y2 in
pscore and escount
"twosteps" replaced by "twostep" in
tobit1
var_ replace by .var in
pscore
new quad_form function
gaze method for mlogit objects
for rsq, type = w, lm, lr changed to
wald, score and lr, same for names in tests for
binomreg, ordreg and
poisreg
argument .vcov of stder is changed to
vcov
gaze.ivreg has now
signif.stars = FALSE
a cmtest method is added for weibreg
objects
new data sets unemp_duration and
recall
ordreg now has a "cloglog" link and a
Surv object can be provided as the response for
right-censored data
weibrug is a new function to fit Weibull models; the
AFT and PH parametrization are provided and the Gamma-Weibull model is
obtained with mixing = TRUE