Readings
--------

IRT models are used for two major purposes:

(1) to assign scores to experimental subjects/examinees/students

(2) to learn about features of subjects or items/tasks that contribute
    to or inhibit correct response

To learn a little about the way IRT models behave in assigning scores
to subjects, please read at least the first 1/3 or so (on the Rasch
model) of the pdf

 * Interactive pdf on Item Response Theory.pdf

There are embedded links to applets that are nice to play with, to
gain some intuition for how IRT models produce scores.  If the
embedded links don't seem to be working, please try the original file
at http://www.metheval.uni-jena.de/irt/VisualIRT.pdf instead.

To learn a little about how IRT models are used to explain task
performance in terms of features of experimental subjects or tasks,
please skim

 * deboeck-wilson-chapter-2.pdf

Unfortunately both readings are a bit technical.  Read around the
technical parts that irritate you, and try to get the main ideas that
the authors are trying to convey.

For those who are curious, there is also a subdirectoring in the
"Readings" directory called "building tests", with a couple of
readings on using IRT models to help build (unidimensional) sets of
items/tasks that will be good at representing performance in a
particular domain of behavior.



Data analysis
-------------

In the "Data" subdirectory there is file 

* stampfer-analysis-part-2.r

with some additional things to try with the stampfer data.  In terms
of the "deboeck-wilson-chapter-2.pdf" reading, these are all "item
explanatory", that is, they attempt to understand task/item
performance in the Stampfer data in terms of features of the tasks or
items (or the conditions under which they were performed).

[note -- this file will not be added to the "Data" subdirectory until
later in the day Weds (after I finish developing it!).]