Fit change-score versus pretest-score data with α and γ process models to diagnose learning processes and distinguish student ability from instruction impact¶

Strategy · All strategies
Evidence · no claims cited

Description¶

The article recommends fitting empirical pre-post change-score data with the theoretical learning curves (α-process straight lines and γ-process logistic curves, alone or combined via Eq. 21/22) to estimate parameters relevant to student learning. A linear change-score relation indicates a predominant α-process with constant α and zero gain–pretest correlation, while a curving-down pattern at low pretest scores suggests γ-type processes or non-constant α. Variable control using additional measures such as reasoning tests or GPA allows separating student ability from instructional effectiveness.

Design Implications¶

Context¶

Requirements¶

• Pre-post score data plotted as change score versus pretest score, ideally with variable control such as holding instruction constant across ability groups
• Additional measures beyond pre/post scores, such as reasoning-ability tests or long-term grades, to operationally separate ability from instruction effects

Constraints¶

• The article states variable control is crucial and extensive results from large numbers of diverse population groups are needed to single out a specific variable's effects
• Predictions can vary when instruction impacts subgroups differently or when subgroup pretest distributions vary

Target Learners¶

• education researchers analyzing pre-post course data

Target Learning Goals¶

• diagnosing learning process types and instructional effectiveness

Affordances¶

• Two Level Dynamic Learning Model Bao

Related Strategies¶

• Learning decomposition: fit exponential learning curves to compare the relative effectiveness of types of pedagogical support

Examples¶

-

Key Sources¶

• Lei Bao. (2007). Dynamic Models of Learning and Education Measurement. https://arxiv.org/abs/0710.1375