NEWS.md

In TSSS: Time Series Analysis with State Space Model

TSSS 1.3.4-5

• Fixed usage of KIND in Fortran file for warning reported in CRAN package check results.

• Specified all variable types explicitly instead of using implicit variable types.

TSSS 1.3.4-4

• Collected legacy Fortran code (DCMPLX, DREAL, DIMAG, DCABS, DCONJG).

TSSS 1.3.4-3

• Added error handling when lag is too large compared to as.integer(period/2) to lsqr().

TSSS 1.3.4-2

• Fixed the LTO (Link-Time Optimization) problems with Fortran source code reported in CRAN package check results.

• Fixed an issue with the S3 method that was reported as NOTE in the pre-test.

TSSS 1.3.4

• Removed C wrapper functions and registered entry points for the routines accessed by the .Fortran interface to call Fortran subroutines.

• URL: http://www.data.jma.go.jp/obd/stats/etrn/ moved to https://www.data.jma.go.jp/obd/stats/etrn/ in Rainfall.Rd and Temperature.Rd.

• Added a new argument log.base = c("e" (default), "10") to season().

• Added error handling when local span is too small compared to max.arorder to lsar().

• Added a ‘NEWS.md’ file.

TSSS 1.3.3

• Added a new argument period (period of one cycle) to lsqr().

• Added a new argument cov (covariance matrix of the smoother) to trend().

• Added a new argument llkhood (log-likelihood) to ngsmth().

• Fixed the number of frequencies for spectrum plots in lsar().

• Fixed a bug in tsmooth() where the value of esterr was not included in the return value.

• Changed grayscale gradient in pfilter() and pfilterNL() plots.

TSSS 1.3.2

• Added examples of pairs plots to the help pages for HAKUSAN and Haibara datasets.

• Fixed a bug in season() and changed the interpretation of argument period.

In the case of period = 7, the data is daily (7 days a week) and not weekly.

• Corrected the source code of the Fortran subroutines FFARMA (armaftf.f) and FFSEAS (seasonf.f) to avoid runtime errors.

• Added error handling for missing data values to arfit(), armafit() and armafit2().

TSSS 1.3.1

• Exported functions (i.e., print.tvvar(), plot.tvvar() and plot.pfilter()) for an extension package TSSSomp using OpenMP.

• Removed the perspective plot from the output of ngsmth().

• Added the following references:

• Kitagawa, G. (2020) Introduction to Time Series Modeling with Applications in R. Chapman & Hall/CRC.

• Kitagawa, G. (2020) Introduction to Time Series Modeling with R (in Japanese). Iwanami Publishing Company.

TSSS 1.3.0

• Added a new argument lag to period().

• Added pfilter() (Particle filtering and smoothing) and pfilterNL() (Particle filtering and nonlinear smoothing) functions.

• Added armafit2() function (Scalar ARMA model fitting: Estimate all ARMA models within the user-specified maximum order using maximum likelihood method).

• Added tsp attribute to all datasets.

• Added Haibara, Nikkei225 and Rainfall datasets.

• Added an example using Haibara dataset in crscor().

• Output the value of arcoef for each order of the AR model in arfit().

• Changed value of noisew of examples in ngsmth().

• Specified the allowed values for trend.order, seasonal.order, ar.order and period arguments in season().

• Added Note about frequency setting, i.e., for data with the tsp attribute, the frequency attribute has priority over the period argument in season().

• Changed the default value for lag from 10 to sqrt(n), where n is the length of the input time series.

• Changed the plot style of evolutionary power spectra in plot.tvspc().

• Changed the function name from armaimp() to armachar().

• Fixed a bug in subroutine istat3 (armafitf.f).

• Changed default values for wminmax in case of noisew = 2 (Pearson) and vminmax in case of noisev = 2 (Pearson) in ngsim().

TSSS 1.2.4

• Used the gammafn function provided by R instead of the DGAMMA function in the original Fortran source, because an error was found in TSSS-00install.out on r-patched-solaris-x86.

TSSS 1.2.3

• Changed to use the warning() function instead of the cat() function to print warning messages.

TSSS 1.2.2

• Fixed Fortran code for subroutine ARYULE (comsub.f) according to the warning message when using gfortran with -fpic -g -O2 -Wall -mtune=native.

TSSS 1.2.1

• Changed the implementation of the random number generator to Mersenne-Twister.

TSSS 1.2.0

• Fixed bugs in arfit(), ngsmth() and trend().

• Fixed Fortran code according to the warning message when using gfortran with -Wall -pedantic.

• Merged gdensty(), cdensty(), pdensty(), exdensty(), chi2densty(), dexdensty() and udensty() into pdfunc().

• Renamed lsar1() and lsar2() functions to lsar() and lsar.chgpt() respectively.

• Added S3 methods for class objects returned by functions.

• Added ‘src/init.c’.

TSSS 1.1.0

• Changed Sunspot dataset into a part of sunspot.year in R`s core datasets package.

• Corrected the tile of the HAKUSAN dataset.

TSSS 1.0.1

• Corrected ‘inst/doc/index.html’ according to the error message from the W3C Markup Validator Service.

TSSS documentation built on Sept. 29, 2023, 9:07 a.m.