-
Notifications
You must be signed in to change notification settings - Fork 1
Home
A Bayesian network is built from a set of variables aranged in an acyclic directed graph. For each variable, the conditional probability distribution of that variable given its parents in the graph. For a discrete network, this takes the form of a conditional probability table (CPT).
If the experts specify the table directly, the number of values to be specified is exponential in the number of parents. For this reason, various parametric forms for the conditional probability tables have been developed, in which the number of values to be specified is usually linear in the number of parents. This package collects R code for generating the CPTs for these various parametric forms.
The second type of tool contained in the package is tools for analyzing the output of Bayesian networks. This includes both graphs for displayig multiple conditional probabilities, and graphs for analyzing sequences of evidence.
The tools here were originally developed to be used with StatShop (a proprietary Bayes net scoring engine developed at ETS), but have been adapted to work with RNetica. There is no dependence on RNetica, the tools here could be easily used (perhaps with some minor adaptation) with other Bayesian network packages.
I recently did a demonstration of CPTtools for my Bayes net class and recorded it using Tegrity. I'm posting a few of those resources here as a way of getting started.
- The Tegrity recording (R transcript not curretly available.)
- The Slides used for the talk. Note that these were revised an based on version 0.2 of CPTtools class and some bug fixes found in the demo are now fixed. (PowerPoint, PDF).
CPTtools is made available under the Artistic License, version 2.0.
The current version (0.6-1) is now fully debugged and contains a users manual. This version also has the new updated discrete partial credit models meant to work with the Peanut parameterized network package.
Also included in the new version are R datasets corresponding to ACED.
New with this release is a method for the lattice function barchart
for Conditional Probabilty Frames. Check it out
The following releases are archived here:
Release
Source
Windows Binary
Mac Binary
Manual
Notes
0.1-4
DiBello-Samejima models only
0.2-3
Weight of evidence graphs merged in.
0.3-1
Discrete Partial Credit framework and complete docs
0.3-2
Peanut Release
0.4-2
Debugging with test cases.
0.4-3
Added normal link and docs back in.
0.5-1
Functions as.CPA and as.CPF moved from RNetica
Added barchart.CPF, requires lattice package.
0.5-3
CPTtools_0.5-3.tar.gz
Added fcKappa and gkLambda functions and fixed a few exports.
0.6-1
CPTtools_0.6-1.tar.gz
Added isMonotonic function..
Note the source versions have been compiled and tested under Linux/Unix, Mac OS X and Windows. As this is written purely in R, no software other than R is required to install this package, and the source package should install on all platforms. (If installing from the R menus, you need to select "Install source package...")
The latest development version is available as a subversion repository at: https://pluto.coe.fsu.edu/svn/common/CPTtools/trunk/.
If you have patches or other correspondence about CPTtools you can contact me at almond@acm.org. If you are filing a bug report, please be sure to include:
- The version number of CPTtools and R that you are using.
- Your operating system and whether it is 64 or 32 bit.
- The exact text of any error message that is generated.
- A small script that reproduces the problem (if at all possible).
The software is offered AS IS, without any warranty of any kind. Support is provided on a volunteer basis, and may not be immediately forthcoming.
Site last updated on 2019-11-21.