Recent Changes - Search:



Bayesian Networks in Educational Assessment

Cognition and Assessment


edit SideBar

BN /


Bayesian Network Manipulation Packages.

There is a fairly long list of these and it is difficult to keep up with the changes. The Bayesian network page at Wikipedia tends to be a good resource for this as the various vendors try to keep the page up to date.

The Bayes net packages below were selected because the book authors have used them at one time or another, and so can recommend them. All Bayes net packages have strengths and weaknesses so if you are doing a serious project, you might want to spend some time checking out the options, including ones not listed here.

  • Netica -- Netica is a commercial Bayes net package that seems to offer a good trade-off between price and availability. A free student version is available which can work with small networks, like most of the ones in the book.
  • RNetica -- A binding of the Netica API into the R environment, which allows R to be used as a scripting language. In addition, R code related to the DiBello-Samejima? models described in Chapter 8 of the book are available in a companion R package called CPTtools. (CPTtools can be used without RNetica). Like Netica, RNetica has a free version which works with small networks; for larger networks a Netica API license must be purchase from Norsys.
  • GeNie and Smile -- GeNie is a Bayesian network package that is free of educational and research purposes. Smile is the corresponding API. GeNie and Smile will read and write the format of many other Bayes net software packages, so they are useful for translating between formats.
  • StatShop? -- StatShop was a suite of tools for using Bayesian network in assessments. It is still experimental software, but a license can be obtained from ETS.
  • Wikipedia page -- The software listing on the Wikipedia page for Bayesian networks is generally up to date.

Many of the popular packages for Bayes net software have graphical interfaces design to run only on windows (the APIs usually run on Linux/Unix/Mac OS). Mac and Linux users might be interested in running windows tools under Unix.

Unusual file extensions on the site.

The examples for "Bayesian Networks in Educational Assessment" where originally prepared in StatShop?, which is a suite of tools for building, calibrating, scoring and experimenting with assessments built using ECD and BayesNets?. StatShop files are all written in XML and use extended extensions to describe the file type:

Graphical Models (Bayes nets)
Assessment Model Description (AssemblyModel?) -- Contains a mapping between task models? and evidence models?, and between tasks? and links?.
Task Data Description -- Lists of tasks with additional data
Calibration Pretest Data -- Actually any data file with student observables, may or may not contain additional demographic data about the students and scores.
Pretest Description -- a file that describes the layout of a comma separated value file, for conversion back and forth between .cpd.xml and .csv format. Many data sets are available in both formats.

Note that in addition to the

Edit - History - Print - Recent Changes - Search
Page last modified on April 16, 2015, at 02:31 PM