Version 1.6 (9 July 2025):
==========================

* garchx(): new value "hac" on the vcov.type argument

* vcov.garchx(): new value "hac" on the vcov.type argument

* new function: confint.garchx()


Version 1.5 (13 September 2022):
================================

* version 1.4 did not pass the CRAN checks

* garchxSim(): new argument 'c.code'

* recursions.c: to avoid the 'ordered comparison of pointer with integer zero' warning, integer comparisons were modified to !=

* new function: refit.garchx(), in response to a suggestion by Viatalie Spinu, see https://github.com/gsucarrat/garchx/issues/3

* class checking in ttest0() and waldtest0() is now done with is()

* minor changes to DESCRIPTION file, and to help-files garchx-package.Rd, garchx.Rd, garchxSim.Rd and ttest.Rd


Version 1.3 (15 July 2021):
===========================

* garchx():
- the non-default value on the backcast.values argument changed, see documentation
- new argument: estimate
- minor changes to the documentation

* garchxObjective(): an additional (positivity) check of the parameters was added; the check is only undertaken when the xreg' argument in garchx() is not NULL

* predict.garchx(): the default of the n.sim argument changed to NULL

* toLatex(): minor changes to the print

* print.garchx(): minor changes to the print

* new functions: gdiff(), glag()

* fix: the returned value from garchxSim(1, verbose=TRUE) is now a matrix


Version 1.2 (23 January 2021):
==============================

* fix: the initial.values argument in garchx() can now be non-NULL


Version 1.1 (10 May 2020):
==========================

* garchx():
- c.code argument: new default TRUE
- new argument: vcov.type
- for consistency with the garch() function in the 'tseries' package, the arch and garch orders changed place in the order argument, i.e. order[1] now refers to the garch order, and order[2] now refers to the arch order
- the 'verbose' argument (w/default TRUE) has been renamed to 'turbo' (w/default FALSE)

* garchxRecursion(): improved speed when the garch order is 1 or greater, since the function now uses compiled c-code for the recursion

* garchxAvar(): new function, computes the asymptotic coefficient-covariance of a model by simulation

* garchxObjective(): minor changes for improved numerical robustness

* vcov.garchx(): new argument vcov.type

* fitted.garchx(): new argument as.zoo (logical)

* residuals.garchx(): new argument as.zoo (logical)

* new functions:
- nobs.garchx()
- predict.garchx()
- quantile.garchx()
- toLatex.garchx()
- ttest0()
- rmnorm()
- waldtest0()

* fixes:
 - garchxSim(): now works when the asym order is greater than 1
 - vcov.garchx(): now works when verbose/turbo=FALSE in the garchx object it is applied to
 - print.garchx(): now works when verbose/turbo=FALSE in the garchx object it is applied to


Version 1.0 (7 April 2020):
===========================

* Package released on CRAN, the version numbering '1.0' means the release is considered to be fairly stable and well-tested