|
Glass Box is a powerful Windows based statistical
package that can be used to fit and analyse complicated non-linear models
and statistical distributions. It is particularly well suited to the
analysis of financial data and extraction of market signals from one or
more indicators. Read the following paper for a discussion of how GlassBox
can be used for this purpose.
Trading With a 't': Signal Extraction Using Non-Parametric
Newey-West Style t-statistics
Features:
- Simple to use
Windows interface.
- Cutting edge
menu-driven statistical modules for fitting high dimensional
relationships, with a minimal set of assumptions. Smooth lines,
surfaces and hypersurfaces can all be estimated so as to capture
conditional mean functions, first derivative functions, standard
errors and multivariate densities.
- Several
post-analysis tools exist for testing model fits, including in-sample
and out-of-sample performance stats, residual histograms, moments and
test for normality, residual and fitted value plots as well as
standard statistical output. These tools are presented as a set of
tabbed pages and controls, which can be quickly navigated with the
mouse.
- Drop down graphing
controls allow for a variety of 2-D, 3-D and 5-D views of model fits
and raw data, including lower-dimensional conditional cross sections,
drill downs and functionals with confidence bounds.
- Wide range of
auxiliary graphing capabilities, including 1-D, 3-D, and 5-D
histograms, line, bar, area, Open-High-Low-Close graphs and more.
Fully customizable graph features, with "smart defaults"
allow for publication quality exporting of graphs with no or minimal
changes, including automatic title and axis labeling.
- The use of basic
Databank objects allows for the easy management of raw data. New data
columns can be created and larger statistical models can be built up
by passing back results from individual statistical modules. Selection
and truncation filters can be easily applied and removed from within a
Databank. Summary stats, correlations and quick graphs are available
with a few mouse clicks.
- Missing values are
automatically handled in every statistical and graphing module. This
allows for the analysis of data with "unbalanced designs",
and the ability to analyze and graph data without having to constantly
pre-filter data series.
- Universal data
import and export facilities, as well as callable DLL versions of
GlassBox allow for an easy interaction between GlassBox and other
applications.
- 5-D graphing
functions allow for the identification of complex patterns and signals
as a function of 4 underlying variables. Several thousand data points
can be graphed as a function of 4 variables, and complex
high-dimensional non-linear patterns can be discerned from such
graphs.
Features Well-Suited to Finance:
- Automatic 2-D graph
handling of dates and times, including the ability to graph signals,
stock returns, open-high-low-close graphs, or any other data set as a
function of the date and time. Graphically drill down to any period of
interest. Up to 36 data series on 12 separate axes can be quickly graphed
via drop down menus, using a single date axis.
- The use of
Newey-West variance estimators for both OLS and non-parametric model
fits, allows for the use of overlapping stock returns data, and other
sources of data subject to both heteroscedasticity and autocorrelation
by correcting the standard errors of the model fits.
- The Signal
back-testing module allows the user to quickly test a variety of
trading strategies (BUY-HOLD-SHORT SELL-TILT) involving a risky asset
and an alternative asset, based on a signal.
- The hit ratio module
allows for robust statistical testing of signals.
- The ability to fit
non-parametric surfaces that "wrap" along periodic edges
makes use of data that is seasonal in nature, or has time of day
effects.
Quick Links:
On-Line Manual (with Screen Shots)
Papers (Adobe PDF Format)
Contact Us
|
