Recommended procedure for applying L* to sequences of affective states¶

Strategy · All strategies
Evidence · no claims cited

Description¶

The article outlines a step-by-step procedure (Algorithm 1) for computing L* on full affect sequences with transitions to A excluded. The algorithm explicitly handles special cases that would produce undefined values or division-by-zero errors, such as sequences consisting solely of transitions of the form Aprev→Bnext, where "L∗ is undefined." After computing L* per student, the authors recommend "a two-tailed t-test can then be used on the entire sample of L∗ values" and a Benjamini-Yekutieli post hoc correction with an α value of 0.05 to control false positives under arbitrary dependence.

Design Implications¶

Context¶

Requirements¶

• Start with full affect sequences including all self-transitions
• Handle undefined cases explicitly (no transitions from A, or sequences solely of AB transitions when using individual base rates)
• Apply Benjamini-Yekutieli correction with α = 0.05 when testing multiple transition pairs

Constraints¶

• When base rates are computed individually, sequences with at most one transition in T_A fall into special cases where L* is undefined or zero

Target Learners¶

• researchers analyzing affect dynamics data from students in digital learning environments

Target Learning Goals¶

• valid analysis of affective state transitions when self-transitions are excluded

Related Strategies¶

Examples¶

-

Key Sources¶

• Jeffrey Matayoshi and Shamya Karumbaiah. (2020). Adjusting the L Statistic when Self-Transitions are Excluded in Affect Dynamics. Journal of Educational Data Mining, Volume 12, No 4. https://jedm.educationaldatamining.org/index.php/JEDM/article/view/478