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¶

Strategy · All strategies
Evidence · 1 claim (1 for) · 1 study (1 associational), q2 · 0 of 1 report an effect size · 1 claim rests on one study

Description¶

The author proposes at least three practical and principled ways to incorporate rule induction into education research so that it complements rather than conflates with regression. First, use ruleset mining to describe the sample and how key independent and dependent variables relate. Second, use association rule mining to identify what factors differ across groups defined by outcome, treatment, or background. Third, use decision trees to examine predictor-outcome relationships after controlling for key covariates, modeling covariates first and then investigating relationships at terminal nodes, as instantiated by algorithms such as the logistic model tree.

Design Implications¶

Context¶

Requirements¶

• Methods must be used in ways that complement rather than conflate rule induction with regression

Constraints¶

• Regression better answers questions about average effects of predictors on the population

Target Learners¶

• education researchers using quantitative methods

Target Learning Goals¶

• designing analyses that leverage rule induction's descriptive and subgroup-detection strengths

Affordances¶

• Rules Vs Rulesets Methodological Account

Claims¶

• A few decision-tree splits identify very high-risk students, such as low-SES students who doubt they will reach college [+W]

Related Strategies¶

Examples¶

-

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