Friends in Uncertainty,
I am pleased to announce two new elicitation tools for continuous
probability distributions created by BayesFusion for the Decision
Analysis and Bayesian networks communities:
(1) Metalog Builder, available at https://metalog.bayesfusion.com/
Metalog distributions (also called Keelin distributions, after their
author, Tom Keelin) are a family of continuous probability distributions
that show remarkable shape flexibility, a choice of bounds, including
infinite bounds, and capability to fit data. They are also efficient in
sample generation, which directly supports environments based on
simulation. Not surprisingly, metalog distributions have been
implemented in leading decision modeling software, such as Analytica.
For more information about the metalog distribution, please look at the
comprehensive article on the topic on Wikipedia
(https://en.wikipedia.org/wiki/Metalog_distribution), the Metalog
Distribution web site (http://metalogdistributions.com/), and the
Metalog Distributions YouTube channel
(https://www.youtube.com/channel/UCyHZ5neKhV1mSsedzDBoqyA). These
sources provide access to articles by Tom Keelin and colleagues on the
topic of the metalog distribution.
The Metalog Builder page mimics the functionality of the Metalog Builder
window in GeNIe, which in turn is inspired by the elicitation tools
built by Tom Keelin and colleagues. It allows for fitting a metalog
distribution to data or to any number of percentiles specified by an
expert. The resulting distributions can be visualized and chosen based
on a compromise between their fit to data and complexity. For more
information about the interface, please refer to the GeNIe Modeler
manual page for the Metalog Builder at
https://support.bayesfusion.com/docs/GeNIe/index.html?equations_metalog-builder.html.
(2) Distribution Visualizer, available at https://prob.bayesfusion.com/
Choosing the right probability distribution over continuous data is a
skill that requires considerable statistical insight. When the
distribution is transformed by an equation expression, the task is
daunting even for an experienced decision analyst.
Distribution Visualizer is an interactive tool for visualizing
expressions with probability distributions through Monte Carlo
simulation. This helps with selecting an expression that is most
appropriate for the task at hand. While the equation may only contain
predefined functions, the set of predefined functions is comprehensive
and there are practically no limitations on the functional form and on
the distributions used. We will love to hear about possible extensions
to the set of functions and distributions. The tool mimics the
functionality of the Distribution Visualizer window in GeNIe. For more
information, please refer to the GeNIe Modeler manual page for the
Distribution Visualizer at
https://support.bayesfusion.com/docs/GeNIe/index.html?equations_distribution-visualizer.html.
Both tools are available to the general public. We hope they will prove
useful to the Decision Analysis and Bayesian network modeling
communities. While we cannot promise immediate implementation of every
suggestion, we will love to hear from you ideas for further extension of
the tools!
Cheers,
Marek
----------------------------------------------------------------------
Marek J. Druzdzel, Ph.D. http://www.bayesfusion.com/
_______________________________________________
uai mailing list
uai@engr.orst.edu
https://it.engineering.oregonstate.edu/mailman/listinfo/uai
