This example was first analyzed in Tatsuoka (1984) using the rule space method. This is both a high quality data set, and Tatsuoka has been generoud with sharing the data, so that it has been analyzed by many authors using many models. In particular, Mislevy (1994) translated the model into a Bayesian network. References and history can be found at the ECD wiki.

The model is based on a cogintive analysis of mixed number subtraction done by Klein et al. (1981). Students were taught by one of two methods: Method A&emdash;convert both mixed numbers into improper fractions,&emdash;and Method B&emdash;subtract the whole number and fractions separately, borrowing from the whole number if needed. The Bayesian network here covers only Method B.

Data Files and Descriptions

We do not have permission to distribute the data, but the directory mixnum has a set of 1000 simulated responses.

Pictures

• Method B Five Skill Proficiency Model (PNG) (SVG)
• Method B Five Skill Proficiency Model Filled In (PNG) (SVG)

Versions of the Models

• mixnumFixedTables (Gzipped Tar file) (mixnumFixedTables/) A version using fixed tables, for scoring not for calibration.
• mixnumTest (Gzipped Tar file) (mixnumTest/) An older version of the Method B model using Ergo rather than Netica files.
• mixnum (Gzipped Tar file) (mixnum/)

Generated on Sun Nov 23 17:49:42 EST 2014.

More information about the file formats and tools to open them can be found at the Tools Page of the ECD Wiki.