[Japanese/English]

R package tvvarOMP

The Institute of Statistical Mathematics

May 21, 2020



1. Introduction

R package tvvarOMP includes two R functions tvarOmp() and tvvarOmp(). tvvarOmp() and tvarOmp() are parallel versions using OpenMP of tvvar() and tvar() in R package TSSS respectively. tvvarOmp() is a function to estimate the variance of time varying AR model and calculates the normalized time-series with almost constant variance. tvarOmp() is a function to estimate parameters of time varying AR model.

Original source code of tvvar() and tvar() were published in Time Series Analysis Programing (in Japanese). Parallel computing with OpenMP may reduce the running time. Functions in tvvarOMP are almost independent of TSSS, except the example of tvarOmp(). If necessary, please install TSSS package.


2. Package installation

We provide several binary files and source file for R on Windows and Linux.

2.1 Installation on Windows

(1) Put tvvarOMP_1.1.2.zip on a suitable directory, then start R (RGui).

(2) From the RGui menu
   Packages ---> Install package(s) from local zip files...
         ---> Select files
         ---> tvvarOMP_1.1.2.zip

(3) From the RGui menu
   Packages ---> Load Package...
         ---> Select one
         ---> tvvarOMP

2.2 Installation on Linux

(1) Put source package tvvarOMP_1.1.2.tar.gz on a suitable directory and

   # R CMD INSTALL tvvarOMP_1.1.2.tar.gz

(3) Start R, then execute R command

   > library(tvvarOMP)

R package tvvarOMP is now available. Once the package is installed, HTML help is available.


3. Test environment

We checked this package on the 64-bit Windows 10 (R-4.0.0) with Intel Core i7-8550U and the 64-bit Ubuntu 14.04 LTS (R-3.6.3) with Intel Core i7 6700HQ .

For the second example in tvvarOmp, the running time with 8 threads on Windows 10 is approximately 2.2 times faster than serial computing. The running time with 8 threads on Ubuntu 14.04 is approximately 1.5 times faster than serial computing.

For the second example in tvarOmp, the running time with 8 threads on Windows 10 is approximately 3 times faster than serial computing. The running time with 8 threads on Ubuntu 14.04 is approximately 1.8 times faster than serial computing.


4. Reference

(1) Kitagawa, G. (2010) Introduction to Time Series Modeling. Chapman & Hall/CRC.
(2) Kitagawa, G. (2005) Introduction to Time Series Analysis. Iwanami Publishing Company (in Japanese).
(3) Kitagawa, G. (1993) FORTRAN 77 Programming for Time Series Analysis. The Iwanami Computer Science Series (in Japanese).



Please send comments and bug reports to ismrp (at) jasp.ism.ac.jp.
This research was partly supported by Function and Induction Research Project held by the Transdisciplinary Research Integration Center at the Research Organization of Information and Sciences, Japan.