.changelog TSA-toolbox
Author: A. Schloegl, 1996-2002 <a.schloegl@ieee.org>


==== 2002 ====

11.04.02 TSA was included in Octave-forge. Changelog not used anymore. 	

05.04.02	AAR		adaptive autoregressive estimation included
21.03.02	MDURLEV, MFILTER	VAR parameters changed from cell arrays to a  Mx[M*p] matrix 
		ARFIT2 		arfit-clone included; demonstrates how to use MDURLEV instead of ARFIT  	
13.03.02	VDURLEV		Bug fixed, VAR output had to be transpsed.
		bstest_VAR	extended ARFIT to test -A too, gives better results
				validation shows that VDURLEV mode=2 is best
07.03.02	VDURLEV		DC output included
27.02.02	LATTICE 	minor bug fixed,
25.02.02	VDURLEV		MDURLEV renamed 
22.02.02  	LATTICE 	minor bug fix
22.02.02  	INVEST1, SINVEST1	minor bugs fix, compatibility to Octave improved
21.02.02  	INVEST1		compatibility to octave
17.02.02  V2.84 MDURLEV 	output PE normalized	
15.02.02	sinvest1 	display of PACF when ARPMX==[AR,RC] fixed 
08.02.02	testsuite/bstest_AR 	shows that lattice 2.83 is better than 2.82
08.02.02  V2.83	lattice modified
08.02.02    	his* moved into ../HISTO/*
07.02.02	quantiles	modified	
07.02.02	hist2pdf, hist2cdf, hist2inv, hist2res2, 

06.02.02 	histomerge	!NEW!
06.02.02	hist2res	SD2 normalized by (N-1) instead of N
04.02.02	lattice		normalization changed from mean to sum 
		sinvest1	confidence interval for pacf (2/sqrt(N)) 
				and ACORF (1/sqrt(N))
03.02.02	sinvest1, invest1, invest0, tsademo
				ACFsd removed
31.01.02 mdurlev	default Mode=5
31.01.02 bstest_VAR  	moved into tsa/testsuite/
			ARfit and MDURLEV evaluated. MDURLEV Mode=5 gave best results, Mode=2 is very close,  
			followed by Mode=1; ARFIT is significantly worse; Mode=0,3 & 4 do not work. 
31.01.02 testsuite	several demos and validation routines are moved into tsa/testsuite/ 
			
29.01.02 mdurlev  V2.82	AR-Polynom including I is returned 
27.01.02 bstest_VAR  	bootstrap validation of mode in MDURLEV
	 mdurlev	NaN's are handled as missing values (using COVM.M)
22.01.02 mdurlev	output argument ACF removed
21.01.02 mdurlev	moved to NaN-tb
20.01.02 vfilter 	tested
19.01.02 vfreqz		spectrum, coherence, directed transfer function etc. implemented
18.01.02 vfreqz		modified
	 mdurlev	muttall-strand and vieira-morf method implemented	   
15.01.02 vfilter  	2nd algorithm of multivariate filter function
14.01.02 vfilter  	multivariate filter function
14.01.02 vfreqz	        multivariate freqz
12.01.02 mdurlev  V2.81	levinson-durbin and lattice do work
	 vardemo	started
03.01.02 mdurlev	Durbin-Levinson-Whittle algorithm: UNDER CONSTRUCTION
03.01.02 acovf		scaling with xcorr corrected 
	 ar		NEW: should resemble AR.M from MathWorks  

==== 2001 ====
29.12.01 mlattice	multivariate Burg algorithm: UNDER CONSTRACTION 
29.12.01 mdurlev	multivariate Durbin-Levinson algorihtm: UNDER CONSTRUCTION

17.12.01 invest1	rms replaced by mean(x.^2)	
         rc2poly, rc2ac, ac2poly, ac2rc, ploy2rc, poly2ac
	                WWW-link corrected
			 	    
15.12.01 acovf		Version 2.76
			Mode 1&2 allowed only if no NaN's		
	
01.12.01 === updated mirrors ===		
08.11.01 lattice	help improved 
05.10.01 quaniles	NEW! help function how to calculate quantiles
01.10.01 plotcdf 	in testing state

21.09.01 flix 		NEW!
	 y2res, hist2res improved (Quantile calc) 	

01.10.01 plotcdf	bug fixed, when more then 1 data column is used, x-scale was incorrect
    Version 2.75  	
30.08.01 y2res		NEW! 
04.05.01 sinvest1	improved
	   histo	display improved
26.04.01 detrend	included, can handle missing values, 
				non-equidistantly sampled data, 
				and can remove higher order polynomial trend
20.04.01 histo.m	Version 2.74		
			bug fixed for case N==1, x =minY;
20.04.01 invest1	Version 2.74		
			Integrated order selection included	
11.04.01 lattice	Version 2.73
			several improvements are made:
			1)	sum (and sumskipnan's) were replaced by mean, this has the effect of
				normalizing with actual number of elements. This seem to improve the estimates
	 		2)	the residual is smaller as for estimates with AR.M
		 	3) 	if NaN/mean is used the lattice algorithm can be used for data with missing values, too. 
05.04.01 acovf		improved estimation for data with missing values. 
			unknown if fft and xcorr methods work correctly for NaN's
03.04.01 arcext			docu improved	    
25.03.01 lattice		replaced nansum by sumskipnan
13.03.01 hist2res		Coefficient_of_variation included
	 histo			bug fixed (scaling of x was incorrect since 08.03.01
 8.03.01 histo			third argin specifies the number of bins
				speed improvement using sparse, casting (UINT16,UINT32) and Octave syntax.
28.02.01 selmo			nansum replaced by sum 
				NC calculation corrected, 
				if more than 1 minimum is found, the smaller model order is returned;
26.02.01 selmo			nanmean replaced by nansum(.)/sum(~isnan(.))
	 sinvest1		errorbar for multiple data series
17.02.01 lattice		made nan-compatible
	 invest0, invest1 	use burg (instead of yule-walker)
	 acovf			option 'unbiased' included, option for estimation method included	
16.02.01 lattice		NEW: burg algorithm
	 sinvest1
	 rc2ar			produces acf, too	 
15.02.01 sinvest1		improved (more features support multiple data series)
14.02.01 sinvest1, tsademo	bug fixed

==== 2000 ====
26.11.00 histo		bug fixed
	 hist2res	Entropy for corresponding Gaussian process included
	 acovf		fft-based algorithm implemented and tested
22.11.00 chaos		chaos simulator                                
25.10.00 selmo, histo	nan's (missing values) considered
28.07.00 selmo, histo,  invest1, sinvest1:	revised
01.05.00 rc2ar		tested (bug fixed)
	 ar2rc		included
07.04.00 durlev, acovf, acorf, parcor, invest0, invest1: almost correction of documentation
	 rc2ar		included
01.04.00 invest1, invest0
	 arcext		included
	 ar_spa		included

==== 1999 ====
20.12.99 acovf, acorf, selmo, invest0, invest1	
			missing values (nan) considered. 
			Unfortunately, not OCTAVE compatible.	
29.11.99 selmo, invest1	criteria for multiple trials included
26.11.99 histo		multiple rows, bug fixed
07.04.99 *	 	References included.
06.04.99 selmo		modified; calculation of BIC corrected 	 
31.03.99 durlev		transformation matrix included
	 invest0, sinvest1, invest1
	 acf2arp, acf2pacf removed
29.03.99 idurlev	inverse Levinson durbin algorithm implemented (including AutoCorr);
	 durlev, parcor	modified	

==== 1998 ====
30.12.98 durlev, acovf,acf,acorf,acf2pacf,arp2pacf,invest1, invest0,sinvest1,earpyw,tsademo
			modified, variable names changed e.g. ACOVF,ACF,PACF,AR,PE -> AutoCov,AutoCorr,PartACF,ARP,E
05.11.98 hist2res	histogram evaluation
19.09.98 invest1	documentation for MOPS included (was missing)
24.08.98 histplot	<NEW>
04.08.98 idurlev	syntax bug fixed (still testing neccessary)

28.07.98 acf		check of input args improved
22.07.98 embed		new
13.07.98 sinvest1	inverse filtering included
	 arp2pacf	minor bug fixed
	 selmo		
05.07.98 acf2pacf 	not needed anymore, 
	 idurlev	instead of acf2pacf for multiple channels
03.07.98 invest0, 
	 invest1, 
	 sinvest1 	modified for multivariate analysis, not OCTAVE compatible because SUM(X,D)
03.07.98 ACOVF		modified, for multiple channels and large lags filter-function used
22.06.98 invest1        ACFsd included
22.06.98 sinvest1       display ACF with errorbar
22.06.98 histo          returns full matrix instead of sparse
20.06.98 acorf		lpq, stderr, qpval included according to the idea of Philip Gray
20.06.98 invest0        ACFsd included
19.06.98 demoperf, durlev, sinvest1, invfdemo, bisdemo	modified for OCTAVE-compatibility
31.05.98 durlev3	inlcuded, modified durlev
19.05.98 pg_autocorr	received from Philip Gray <pgray@agsm.unsw.edu.au>, 
			I consider to include it.
17.05.98 Compatibility with Octave tested
17.05.98 histo		modified: with Matlab, sparse function is used. 
17.05.98 demoperf	modified
17.05.98 parcor		included
14.05.98 ucp & hup	renamed to old/*.232
	 ucp2 & hup2	renamed to ucp and hup
	 tsademo	modified (bug fixed)
	 invest1	modified (bug fixed)	 
	 invest0	modified (bug fixed)
	 invfdemo	modified (bug fixed)
	 demoperf	modified (bug fixed)
12.05.98 yuwa 		modified 
27.04.98 durlev		included <new>
27.04.98 acovf		multiple channels
27.04.98 earpyw		multiple channels
27.04.98 selmo		multiple channels
27.04.98 arp2pacf	prepared for multiple channels	 
25.04.98 demoperf, 
	 selmo		minor modifications for Octave compatibility
23.04.98 hup2		(multiple) test for Hurwitz polynomial, included <new> 	 
22.04.98 ucp2		(multiple) test for Unit_Circle polynomial, included
20.04.98 hup & ucp	test for Unit_Circle and Hurwitz polynomial, included <new>	 
30.03.98 arp2pacf	included <new>	 
21.03.98 		LGPL 
	 acovf		modified 
14.02.98 bisdemo	octave combatibility
	 demo		octave combatibility
	 yuwa           octave
         demoperf       ocatve
22.01.98 		minor modifications for compatibility to Octave

==== 1997 ====
14.10.97 sinvest1	modified for Matlab5.0
20.08.97 hiocum		included and removed (only 2-dim matrices)
19.08.97 tsademo2	removed
16.08.97 sinvest0	calc. of Poles corrected
04.08.97 acf2pacf	calc. of PE power of error process included
         invest0
07.07.97 Major revision of TSA Toolbox
04.07.97 histo, entropy	histogram and entropy function included.
20.05.97 earpyw		Remarks modified
  			flops(0); ar(eeg8s,10);flops
			ans = 628013
			flops(0); earpyw(eeg8s,10);flops
                        ans = 23386
12.05.97 sinvest1	poles and bode included,
	 menu  modified
	 invest1	modified
22.04.97 sim2 	removed
18.04.97 sim2		included
	 tsademo2       included
28.01.97 sinvest1	spectrum from 0..1
24.01.97 bispec 	improved (2d rectangular window) 
22.01.97 invest1: 
	 calculation of bicov removed
	 biacov removed
	 bispec removed
22.01.97 biacov added
	 bispec added
	 bisdemo added
21.01.97 biacovf:	bug fixed, made calculation more efficient
 	 bispec		(testversion 0.10) included 

30.11.96 sinvest1: Matched Filter;
 	 eeg8s DC removed 
29.11.96 sinvest1: Determination of Regression instead of Ratio of Determination
19.11.96 sinvest, invest0, invest1: minor errors removed

27.09.96 earpk	input parameter changed
18.09.96 invest0, invest1, selmo, sinvest1:
			Include BIC, SBC, MDL, CAT criteria
	 eigval:	included
	 yuwa:		fast levdur algorithm inlcuded
	 levdur:	removed
04.09.96 acf2pacf: 	returns ARPmx
	 invest1: 	Improvement calculating E(p) using ARP-Matrix
21.08.96 earplms: 	included
19.08.96 invest1: 	Help corrected
19.08.96 sinvest1: 	bug fixed, display R
13.08.96 invest1 & sinvest1: 	MDL & PHI criteria included
30.07.96 sinvest1: 	corrected
25.07.96 invest1: 	Error curve corrected, NO AR parameters returned
17.04.96 invest1: 	bug fixed
12.04.96 TSA toolbox extracted from TARPE toolbox
	 First version published at MathWorks