2014-01-25  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.3-1.
	* Patched to work with recent changes to Rcpp.

2014-01-01  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.3-0.
	* Reversed log(0) fix in VaRTest function (produced wrong answer).
	* Added realized GARCH (realGARCH) model of Hansen, Huang and Shek (2011),
	with almost complete methods.
	* Added new library in Imports (expm). Matrix power (%^%) required for the 
	multi-period realGARCH forecast).
	* Expanded multi-methods to deal with realGARCH and mcsGARCH model.
	
	
2013-11-28  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.2-9.
	* Fixed the stationarity condition of the E-GARCH model for the case when p>1,
	and changed default lower bound on beta to be -1.
	* Fix to the likelihood summation to take into account the first value of the 
	standardized residual (was previously set to zero except for the fGARCH model).
	

2013-08-30  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.2-8.
	* small correction to quantile function for ugarchsim/ugarchpath (resulted
	in error when called as there was a stray comma in the matrix size 
	setup). Fix also to quantile method for roll (resulted in error when called).
	* correction to calculation of hessian after previous version removal of Rcpp
	code introduced an old bug.
	* Added SkewHyperbolic import for the use of the quantile and distribution 
	functions in R.

2013-07-07  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.2-7.
	* removed Rcpp hessian calculation code (caused memory leak), and reverted
	to pure R code. Timings indicate little to no increase in timing.
	
2013-06-22  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.2-6.
	* Change defaults for switching between C and Cpp code for simulation, for 
	faster execution (with implications for rmgarch simulation methods).
	* Small correction in censoring calculation of the VaR Duration Test
	[thanks to David Ardia].

2013-05-26  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.2-5.
	* Major cleanup of distribution code (C source and R). Most R distribution
	functions linked to underlying C code for speedup (affects distfit and 
	related methods). Also reversed automatic vectorization of dpqr	functions
	via vectorize function (created too much overhead), and implemented custom
	solution.
	* Fix to rgig C code.
	* Standardized residuals plot now makes use of residuals method to be 
	consistent with other methods (previously used 'z' value from likelihood
	which has zeros in the initial startup values).

2013-04-24  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.2-4.
	* Small fix to ugarchsim for csGARCH model. The option for using startMethod 
	'unconditional' did not apply to the permanent component of variance and used
	instead the sample based value which led to a small discrepancy between the 
	results run against an equivalent ugarchpath formulation [ thanks to P.B ].
	* Small fix to uGARCHdistribution bivariate plots when no ARMA was used.
	* Re-wrote the C code likelihood for the NIG to use the simplified version
	rather than calling the GH distribution with lambda=-0.5, slightly improving
	speed.
	* Fix to starting parameters of eGARCH model (alpha set too high causing problems
	when using variance targeting).
	* Fix to lower/upper bounds of constant (mu) parameter of arfimafit to catch
	cases where the mean is zero.
	

2013-04-12  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.2-3.
	* Changed calculation of diurnal component of mcsGARCH to use the median 
	instead of the mean for robustness purposes given the type and size of data
	used. Small fix in forecast function for mcsGARCH when completely out of 
	sample forecast requested (resulted in an error).
	* makevars and makevars.win written using portable (backticks) representation 
	and therefore removed GNU make requirement from DESCRIPTION.

2013-04-06  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.2-2.
	* Version number bump for upload to CRAN for R-3.0.0.
	* Small fix to fpm (failed for some timezones-related to as.POSIXct use and 
	internal represenation of time/dates).
	* Small fix to mcsGARCH model again related to timezones and use of 
	as.POSIXct function.
	* Updated some of the tests in the rugarch.tests folder (and moved folder to
	main directory). Added the folder to the .Rinstignore file to avoid 
	installation [suggestion of B.R.].
	* Moved 'Requires' packages to 'Imports' in DESCRIPTION file and 
	more extensive use of importFrom in NAMESPACE [suggestion of B.R.]. 

2013-03-22  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.01-6.
	* Added the multiplicative component GARCH (mcsGARCH) model of Engle and 
	Sokalska (2012) for intraday GARCH modelling (takes into account the
	intraday seasonality component of the absolute returns). Requires regularly
	sampled intraday data (e.g. 1min, 5min etc). Vignette updated with model 
	details.
 	
2013-02-22  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.01-5.
	* Added a "quantile" and "pit" method for certain objects to return the 
	conditional quantiles and probability integral transform (cdf transformation).
	* Added the skdomain function/plot to graph the skewness-kurtosis authorized
	domain of the skewed and shaped distributions.
	* Added a "reduce" method to automatically zero (and exclude) non-significant 
	parameters and re-estimate a model.

2013-02-04  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.01-4.
	* Added/Exported a couple of Date/Time utility functions (move and generatefwd) 
	for working with the returned forecast objects. Examples are in the help file.
	* Small fix to fixed parameter return vector in post-estimation (when both 
	fixed parameters and variance targeting used, the coef method did not output
	the omega value creating problems for other methods which depended on this).
	* Small fix to GMMTest when skew and shape are time varying (does not affect
	GARCH models). Updated documentation on returned object (was outdated).
	* HYP likelihood now handled separately from GHYP for speed in C code (i.e.
	when choosing GHYP with fixed par for ghlambda of 1).

2013-01-30  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.01-3.
	* Methods 'sigma' and 'fitted' are now uniformly implemented for most uGARCH 
	classes for extracting the conditional sigma and mean from all objects. The 
	as.data.frame method is now deprecated. In particular, for the uGARCHforecast 
	object, these methods now return the n.ahead by (n.roll+1) matrix of 
	conditional forecasts where the column names are the T[0] time index prior 
	to the forecast, and the row names labelled as T+1, T+2, to T+n.ahead. The 
	as.data.frame still remains for some objects which do not return regular 
	exclusive time series objects (e.g. uGARCHdistribution, uGARCHroll, and 
	uGARCHboot).
	* xts is now the only supported data "format" accepted by the package. Numeric,
	data.frame and matrix data which does not have a proper time based index is
	coerced into one by assuming the numerical index of the supplied data represents
	number of days since 1970-01-01 (the warning is turned off at present).
	* Removed some functions, namely ForwardDates and WeekDayDummy, since the move
	to xts makes them obsolete.
	* Removed the as.array and as.list extractor functions in the multi-function 
	object classes (replaced by sigma and fitted).
	* Consolidated code on methods (using mostly 'switch' functionality), to 
	eliminate some of the code redundancy.
	* Added VaRplot function for plotting the VaR exceedances.
	* Added VaRloss function for calculating and returning the VaR loss used in
	Gonzalez-Rivera, Lee, and Mishra (2004).
	* Fix to ugarchboot when using a uGARCHspec object with the csGARCH model.
	The fix allows to pass the previous permanent component of the variance
	(preq) in the ugarchpath method as an additional argument via the dots(...).
	The vignette now includes a step by step recipe of how the full bootstrap is
	constructed.
	* Turned off warning about NaN's in post-estimation numerical derivative 
	calculation and added a note in the ugarchfit-methods explaining its purpose
	and how it relates to the optimal solution.
	* ugarchroll now also flags as a non-converged estimation windows one where
	the Hessian cannot be inverted (indicating a local solution).

	
2013-01-24  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.01-2.
	* Distribution functions re-written to take advantage of vectorization (some
	of them were not vectorized), and where deemed appropriate, parallel evaluation
	(for non-analytic quantile/distribution functions e.g. qghst). This is needed
	in particular for another package in development. Functions ddist, qdist, 
	pdist, rdist, dskewness and dkurtosis now take advantage of location and 
	scaling property of standardized NIG/GH distributions for faster evaluation.
	* rugarch-distributions.R file housekeeping. Elimination of redundant code.
	* Some cleanup to plots (now return old par setup on exit).
	* Analytic skewness and kurtosis expressions for Fernandez and Steel's sstd
	distribution.
	* Small fix to ugarchroll resume method.

2013-01-18  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.01-1.
	* Version numbering bump.
	* Analytic skewness and kurtosis expressions for Johnson's SU 
	distribution.
	* distplot function provides skewness-kurtosis plots (surfaces) for 
	skew-shape combinations of different distributions.
	* Corrected some typos in documentation/vignette [thanks to D.E].

2013-01-15  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.0-16.
	* Fix to bug in autoarfima for method partial (convergence check
	failed)
	* Change initialization of arfima process for values < armaOrder to a
	constant.
	* Added extra M(1,2) and M(2,1) to the Hong and Li test (HLTest) report 
	which are related to ARCH in mean and leverage effects respectively.
	* Fix in some methods in the presence of variance targeting.
	* Fix to some simulation methods when using external regressors in variance
	equation and unconditional choice for starting the recursion.


2012-11-30  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.0-14.
	* Removed use of custom function environments in the estimation. Passing
	instead an argument list between functions (resolves various problems for
	non-shared environments e.g. use of snow and certain solvers). 
	* Parallel estimation now done entirely using the parallel package, given a
	user supplied cluster object.
	* ugarchroll and arfimaroll completely re-written and some functionality 
	changed (see help files). as.uGARCHforecast and as.ARFIMAforecast methods 
	removed, and access methods on the rolling objects simplified, as the object 
	returns a much cleaner output. Additionally, a new method called resume will
	allow the resumption of rolling objects which had non-converged windows using
	different solver and control option combinations. All methods on objects
	with non-converged windows will now stop and warn. A new option window.size
	allows to control the amount of data used for the moving window estimation.
	* ugarchspec now warns on misspelled options in the specification. Exact 
	(not partial) matching is done on the options in the variance and mean model
	lists [thanks to Pat Burns].
	* variance.targeting option in GARCH specification now allows either logical 
	or numeric value (in which case that is used instead of the unconditional 
	variance of the conditional mean residuals) [user request].
	* fit.control takes an extra option, 'rec.init' for the type of method used 
	to initialize the GARCH recursion ("all" for all data, integer value >1 for
	using the first n points, or positive numeric value <=1 for exponential
	backcast method)
	* The GARCH bootstrap method (ugarchboot) has been re-written and enhanced
	to allow for sampling from both the raw residuals, kernel fitted residuals
	or SPD (new dependency with spd package) fitted residuals. Returned objects
	are also more simple to manipulate with extractor methods.
	* Fix to GHST distribution density.
	* Fix to arfima simulation initialization requirement (n.start>=MA order).
	* Minor code fixes
	* vignette updated
	
2012-10-30  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.0-13.
	* Fix to hybrid solver default trace argument
	* 2 sided Hessian now moved to C++
	* Some enhancements/fixes to date identification utility for recognising
	date formats
	* Fix to external regressors in variance equation (bug when more than
	2 regressors used) [thanks to Florian Ziel].
	* Fix to starting and fixed parameters for high order models (>20).
	* Ability to set upper and lower bounds via setbounds<- function on
	a  uGARCHspec object.
	* GARCH starting parameters functions cleanup/compacting.
	* Some fixes to show method for very high dimensional models. Also, some 
	changes to Box Test summary and Information Criteria.
	* Fix to arfimadistribution function.

2012-09-30  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.0-12.
	* Added a hybrid strategy optimization method to sequentially try a list of 
	alternative solvers in case of non-convergence.
	* Added optional logical argument 'standardize' (default: FALSE)
	to residuals method to return the standardized residuals for both the
	arfima and garch classes.

2012-07-09  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.0-11.
	* Added the Component GARCH model of Lee and Engle (1999)
	with a complete set of methods.
	* ugarchroll now only allows n.ahead=1 (alternative was too complicated).
	* Change to ugarchboot to allow simulated external regressors input 
	separately from the forecast of the external regressors.
	
2012-06-30  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.0-10.
	* Fix in ugarchboot (full method) for higher GARCH and ARMA orders.
	* Numerous fixes to methods involving variance targeting.
	* Fix to egarchsim when n.sim<100 and m.sim>100 (c++ code -- brackets in 
	wrong place!)
	* Fix to ugarchroll using multicore, and correct f01density slot.

2012-05-23  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.0-9.
	* Added autoarfima function to select best arfima model based on information 
	criterion using optionally complete enumeration of the combination space for 
	the max AR and MA orders chosen.
	* Fix to some .C calls to eliminate Null being passed (affected only filter 
	method).
	* Fix avoid error in uncertainty functions for arfima and garch 
	(ugarchdistribution and arfimadistribution) when using spec rather than fit.
	* Bounds for AR and MA increased to [-4,4] for models with orders higher 
	than 1.
	* Fix in ugarchroll to include external.regressors in forecast

2012-02-03  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.0-8.
	* Turned off stationarity check during calculation of numerical derivatives
	to avoid boundary case problems.
	* Addded nloptr solver to estimation toolbox.
	* Added Generalized Hyperbolic Skew Student distribution ("GHST").
	* Added a battery of tests including VaRTest, VaRDurTest, ESTest, GMMTest
	and HLTest.
	* Renamed BerkowitzLR to BerkowitzTest.
	* Updated Vignette with details on changes.
	* Fixed a bug when spec had include.mean = FALSE but fitting routine 
	was still subtracting the mean.

2011-12-16  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.0-7.
	* Fix to eGARCH model in the presence of variance.targeting.
	* Fix to ugarchspec when external.regressors in variance equation > 1.
	* Vignette clarifications.
	* Added (experimental) archex argument in mean.model [user request].
	* Added vcov and convergence extractor methods to uGARCHfit object.

2011-10-17  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.0-6.
	* '&' replaced by 'and' in some plots and summary methods for
	better compatibility with tikz (requested by a user).

2011-10-15  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.0-5.
	* Fix to makevars (for solaris build).
	* Fix to avoid multicore note on windows checks.

2011-10-12  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.0-4.
	* Added tail.test option to BerkowitzLR test. This is a likelihood ratio 
	test based on censored normal for testing fit of the tail using the 
	probability integral transform (PIT).
	* Added an example to the DACtest function (in documentation).
	* Fixed a bug in rolling forecast when calling ghyp functions with variable 
	length lambda (changes to qdensity and scaledist to handle vector inputs of 
	lambda).
	
2011-09-28  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.0-3.
	* Fix to starting and fixed parameters for AR coefficient.
	* Added a show method for uGARCHroll.
	* Fixed an issue with passing a data.frame object with numerical
	indices as rownames (affected most ugarchroll extractor methods).


2011-09-20  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.0-2.
	* More minor changes to some show methods and ugarchbench.
	* Changes to optimization limits for external regressors.
	* Bug fix in rolling forecast with arfima [thanks to John Kerpel].

2011-09-05  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.0-1.
	* Some minor changes to reduce warnings on CRAN check.
	* packageStartupMessage used for .onLoad
	* Export of GARCHfilter, uGARCHfilter, GARCHroll and uGARCHroll classes 
	(forgot them on original release!).
	* Optional use of multicore will still generate warnings (since not 
	available for Windows).
	* Vignette pdf is pre-built (Rnw simply links to the pdf).
		
2011-08-30  Alexios Ghalanos <alexios@4dscape.com>
	* DESCRIPTION (Version): New version is 1.0.
	* rugarch represents an almost complete re-write of the univariate GARCH 
	models of the rgarch package which is now being split into seperate 
	univariate (rugarch) and multivariate (rmgarch) packages for easier 
	maintenance.
	* Completed Vignette.
	* Numerous corrections, elimination of redundant code, and some new 
	functionality has been added.
	* ARFIMA simulation corrections and multiple checks.
	* rugarch.tests folder now contains an almost comprehensive set of examples 
	and tests which can be run via a wrapper function found in the runtests.R 
	file.
	* First upload to CRAN.