Title: Extending the pks Package for Fitting Probabilistic Knowledge
Structures

Author: Florian Wickelmaier

Abstract:
The pks package in R implements parameter estimation and response pattern
generation for the basic local independence model (BLIM), one of the most
popular probabilistic knowledge structures. A knowledge structure is a family
of sets, called the knowledge states, each of which contains the items a
student of some domain is mastering. Recent additions to the package aim at
enhancing its use for analyzing response patterns. First, item tree analysis
is a data-driven method that seeks to establish a precedence relation among a
set of binary items. From the precedence relation, a quasi-ordinal knowledge
space is obtained. Second, a challenge in fitting the BLIM is its potentially
large number of knowledge state parameters. The simple learning model (SLM)
restricts the state parameters by introducing a learning process on the items,
thereby reducing the number of state parameters. Finally, plotting facilities
for knowledge structures help visualizing dependencies among items. The
presentation illustrates some of the existing and recently added features of
pks with real-world data sets.