An Open Source Time Series Analysis Package
Latest Version: 1.41
- Graphical User Interface
- Highly Efficient Numerical Algorithms
- Estimates and Forecasts for
- ARMA/ARIMA models
- Long-memory ARFIMA models
- ARCH/GARCH/EGARCH models
- Stochastic Volatility Models
- Handles missing data
- Both Bayesian and Frequentist Analysis
- Freely available with source (GNU General Public License)
(Click either to install or to upgrade from an older version.)
Note: Under the terms of the GPL, this program is free software; you can redistribute it and/or
modify it under the terms of the GNU Library General Public
License as published by the Free Software Foundation; either
version 2 of the License, or (at your option) any later version.
This library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Library General Public License for more details.
The file COPYING in the source package (.tar.gz file) below contains the
GNU Library General Public License.
| Version || Date Posted ||Windows Executable || Complete Source || Notes |
| 1.41 || Mar. 23rd, 2008
- bug fixes for new gtkmm library: screen refresh, eps export
| 1.40 || Jan. 17th, 2007
- several aesthetic improvements
- forecasting is now Monte Carlo (simulation) based -
this allows for accurate estimation of predictive quantiles
for nonlinear and non-Gaussian models
| 1.31 || Jul. 10th, 2006
- bug fix release: eliminated intermittent crashing problem after specifying ARMA models
| 1.30 || Apr. 27th, 2006
- supports EGARCH models
- more bug fixes
- time-reversal transformation added
| 1.27 || Mar. 22nd, 2006
- more bug fixes
- "custom" transformation added: can enter standard functional expressions
| 1.25 || Feb. 17th, 2006
- more bug fixes, package name change to Cronos
- Windows version has a standard Windows installation package
- spectral densities and periodograms included
- can view continuous-time (diffusion) versions of models
| 1.21 || Apr. 8th, 2005
- Same gtk+ and gtkmm runtime required as for 1.10
- Changes: residual windows include q-q plots instead of PACF
| 1.20 || Mar. 1st, 2005
- Changes: stochastic volatility models supported + bug fixes + scaling of ACF/PACF plots
| 1.10 || Feb. 23rd, 2005
- Changes: bug fixes + computation of standard errors of param. estimates
| 1.01 || Sep. 21st, 2004
- Uses gtk+ runtime 2.4.5, gtkmm runtime 2.4.9
This is a time series analysis package written using the Gtk+ graphical
toolkit, with the following features.
From the point of view of programmers/researchers, it has the following
- It is open-source (you can download the source and/or selected
binaries below), under
the GNU General Public License.
- It has a nice graphical interface (see screenshot below), based
on the Gtkmm C++ interface for the
GTK+ graphical user interface library.
- It carries out parameter estimation and forecasting
for ARMA/ARIMA/ARFIMA, GARCH and SVM models.
- Both Bayesian and frequentist estimation methods (MCMC, MLE) can be used.
- It handles missing values.
- It allows graphics to be exported to encapsulated postscript files,
for inclusion in documents created, for instance, with
- It makes use of the GNU GSL (GNU scientific library, also released under the GNU General Public License. ) and
libraries, which contain highly-efficient code for carrying out
optimization, solving systems of linear equations, etc.
- Portability: Both Windows and Linux versions are currently available, and since it's open source, in theory, it should be compilable under other operating systems as well.
- Extendability: The code is object-oriented C++ code. In particular, there is a class named "TimeSeriesModel", and ARMA, GARCH and SVM model structures are derived from this. It is relatively straightforward to derive a new class from the base "TimeSeriesModel" class and integrate it into the existing framework.
Download and execute the setup file.
(You can also just click on the large button above.)
If you do not already have them, you will be
asked to install the gtkmm
run-time libraries on your system.
When you install these, you may be asked to
install gtk+ run-time libraries on your system.
Answer yes to both questions. Without these run-time
libraries, Cronos will not work.
A shortcut will be placed on the desktop during installation.
For Linux (Compilation Required)
Prerequisites: You'll need the development version of
the Gtkmm-2.4 (or later) library, as well as
the development version of the Gtk+-2.4 (or later) library.
In addition, you'll need the GNU scientific library version 1.4
After making sure the development packages listed above are installed on your system,
download the Cronos source tar file (below),
untar it into a directory, go into the directory, and at the command prompt, execute the following commands.
The configure program takes the usual autoconf options, see the INSTALL
file contained in the source package (.tar.gz file) for more details.
- make install
Note that I haven't quite worked out how to get the automake tools to
ensure that linking specifies libraries in the correct order.
Yo may have to link manually if you get unresolved references.
Windows Compilation (Not necessary unless you want to customize/alter the program in Windows.)
In Windows, you can also try to compile Cronos directly from the source package.
This isn't particularly easy, but, roughly speaking, can be done
using the following steps.
- Install MinGW , including MSYS and the GCC compiler.
- Download and install the development versions of the gtk+ and gtkmm libraries
(same links as given above.)
- Download and compile (using g77 from within MSYS) the BLAS
and LAPACK libraries.
- Download and install the GNU scientific library (gsl), version 1.6 or later, from within MSYS.
Download the Cronos source (below), start MSYS, and follow instructions for
Linux installation above.
Unfortunately, this has not yet been written. However,
the program is (hopefully) somewhat logically organized, and easy to work out how to
The fundamental principle of operation is that, at any given time, the program keeps
track of two things: a time series data set, and a time series model. All operations
apply either to the current data set, the current model, or (for instance, in the case
of residual analysis) to the combination of the two.
To work with graphs (i.e. change properties, export to encapsulated postscript files, etc.),
right-click the mouse over them.
If you have general comments on the usefulness of the program, questions about making it work, bugs to report, etc., feel free
to contact the author, Anthony Brockwell.
Last modified: Jan. 17th, 2007, by Anthony Brockwell.