Rules-versus-rulesets methodological account: rulesets answer exploratory sequential questions while regression answers average unique-impact questions¶

Theory · All theories
Evidence · no claims cited

Description¶

This account distinguishes what rule induction methods can and cannot answer relative to regression. Ruleset induction "is better suited for modeling rather than identifying specific rules" because rules lose relational meaning outside their ruleset context, whereas association rule mining conducts an exhaustive search producing rules independent of the algorithm and other rules. Ruleset induction answers an exploratory, sequential question about which sets of characteristics characterize each outcome level, while regression answers how much unique impact each independent variable has assuming equal impact across individuals. The author used this account to position the methods as complements rather than substitutes.

Design Implications¶

Context¶

Requirements¶

• Researchers must clearly understand the difference between mining rules and mining rulesets and the unique research questions each answers

Constraints¶

• Ruleset rules are functions of the search algorithm and other rules in the ruleset, making individual rules less reliable
• Rule induction is not directly helpful for the research questions regression is specifically designed to answer

Target Learners¶

• education researchers using quantitative methods

Target Learning Objectives¶

• choosing appropriate analytic methods for predictor-outcome research questions

Claims¶

Related Theories¶

Examples¶

• Use rule induction in three complementary roles: ruleset mining for sample description, association rule mining for group differences, and decision trees for controlled predictor-outcome checks

Key Sources¶

• Iwatani, E. (2019). How Rule Induction Data Mining Can and Cannot Be Useful for Education Research. Paper presented at the 2019 annual meeting of the American Educational Research Association. https://doi.org/10.302/1438441