Robust
Analysis
Inc
Robust Analysis, Inc. is a consulting and software company that
specializes in analyzing problems that involve heavy tailed
distributions. Our initial focus has been on stable
distributions, a class of probability distributions that generalize
normal distributions and allow heavy tails and skewness.
This rich class of probability distributions can model data
in a wide range of applications. In engineering, noise in some
communications, radar and sonar systems is heavy tailed.
In economics and finance, stable laws are
used to model fluctuations of stocks prices and foreign
exchange rates, compute Value At Risk, and model large insurance claims.
They are used in hydrology to model super-diffusion
of a fluid through an aquifier. Also some network
traffic quantities, e.g. file size of downloads, are heavy tailed.
Stable densities computed with the STABLE program.
The values of α are indicated on the plots, skewness is indicated by color:
β=0 (black), β=0.25 (red), β=0.5 (green), β=0.75 (yellow), β=1 (blue).
Basic information on stable
distributions
Basic facts about stable distributions can be found in an
introduction to stable distributions (PDF format, 360 kb). This
is a draft version of Chapter 1 in a forthcoming book on stable
distributions. It describes one dimensional stable distributions in
a non-technical way, explaining the two main parameterizations of
stable laws.
John Nolan's academic webpage has
more information about stable distributions and a free Windows
programs for doing basic calculatins with stable distributions.
STABLE library and interfaces
We have developed a library of STABLE functions for programmers to call directly (from C,
Fortran, or Visual Basic) and interfaces for matlab, S-Plus,
R, and Mathematica. All the capabilities of the STABLE program are available, plus
more features, at increased speed. The library has entry points to compute stable
densities, cumulative distribution function, quantiles, hazard
function, etc. at an arbitrary vector of values. It will estimate
stable parameters by six methods: quantile method, empirical
characteristic function method, numerical maximum likelihood
estimation, fractional moments, log absolute moment, and modified
quantile. Linear regression with stable error terms is now included.
New features include support for bivariate stable distributions:
density calculations, cdf (probability of being in a rectangle),
simulation and estimation of spectral measure. For an
isotropic/radially symmetric distribution X, the amplitude is
defined by R=|X|. This is a generalized Rayleigh distribution.
Functions are included to calculate the density
and cumulative distribution of the amplitude distribution, as well
as simulating samples from the amplitude distribution.
The user manual contains a description of all the functions, see online versions below.
Here is a one page Power Point information sheet on STABLE.
This library and/or interfaces for matlab, S-Plus,
R, and Mathematica can be purchased below.
STABLE version 5.1 was released in January 2009. It is completely rewritten
and now runs under both Windows and linux.
In July 2009 we released the α-stable signal filtering toolbox for matlab.
This toolbox implements nonlinear stable filters for filtering out
spikey/impulsive/heavy tailed noise. The filters can be unweighted or
weighted. In addition, signed weighted and matched filters are defined
for the detection of waveforms, e.g. a chirp, in a noisy signal.
Purchase STABLE online
The STABLE package can be ordered online using a credit card. Credit card orders are usually shipped the next working day.
To order,
look down the rows of the table below for the version you want.
Note that for matlab, you can purchase just the STABLE core toolbox (first row),
or you can purchase the combined STABLE core and the signal filtering (second row).
Then go over
to the column for Windows or Linux and click on the icon that says "E-commerce By Reg.Net".
This will link to a secure server (RegNet.com), that will process your order verify your
credit card. After that is done, they will send us your order and we will ship the software
to you by e-mail.
| Version | cost | User Manual | 32 bit Windows platform | Linux platform |
|---|
matlab STABLE toolbox (no signal filtering)
(matlab version 7.1 or above required)
 |
$300 | STABLE for matlab |
 |
32 bit linux only
|
matlab combined STABLE and signal filtering toolbox
(matlab version 7.1 or above required) | $500 |
STABLE and Signal
Filtering for matlab |
|
Future development |
Upgrade to signal filtering toolbox; you must already have
purchased the base STABLE toolbox
(matlab version 7.1 or above required) | $300 | see above |
|
Future development |
| R | $300 | STABLE for R |
|
Future development |
| Mathematica | $300 | STABLE for Mathematica |
|
limited support - contact us |
| Stand-alone C library (binary) | $300 | STABLE for C |
DLL for Windows
|
.so for 32-bit and 64-bit
limited support - contact us |
| Excel Add-on | | |
Under development | |
| Prearranged software/consulting payments | contact us | |
|
|
For more information about other environments, volume pricing (more than 2 copies),
to order by mail, or other issues, use the contact page listed below.
Grant program
Note: this grant program is officially over,
but a limited number of grants may still be made.
Academic and non-profit organizations may apply for a grant program that
awards a limited number of free copies of STABLE. Decisions on grant
applications will be based on concrete plans for use of the STABLE program and for disseminating results
in publications, talks, etc.
Click STABLE Software Grant Application
for more information.
More information, mail orders, other platforms and consulting
If you have any questions about the above products, are interested in a custom
interface for STABLE to a different program or on a different
platform, or other consulting work, please contact us using our
contact page
Page visits since 29 August 2010: 
