Table 3.
Best model includes the interaction between SES status and the percentage of constructed-response questions on an exama
| Rank | Modelb | AICc | ΔAICc | ωi |
|---|---|---|---|---|
| 1 | Cum.GPA + Time + Gender + SES + W.Diff + Percent CR + SES*Percent CR + Course | −41285.44 | 0.00 | 0.61 |
| 2 | Cum.GPA + Time + Gender + SES + W.Diff + Percent CR + SES*Percent CR + Gender*Percent CR + Course | −41283.64 | 1.81 | 0.25 |
| 3 | Cum.GPA + Time + Gender + SES + W.Diff + Percent CR + SES*Percent CR | −41281.96 | 3.49 | 0.11 |
| 4 | Cum.GPA + Time + Gender + SES + W.Diff + Percent CR + SES*Percent CR + Gender*Percent CR | −41280.15 | 5.29 | 0.03 |
| 5 | Cum.GPA + Time + Gender + SES + W.Diff + Percent CR + Course | −41268.62 | 16.82 | 0.00 |
aRelative ranking (from most support to least) of the four best models for predicting student exam performance using AICc model selection. Models that are informative (ΔAICc < 10) are shown, plus the next best model that had a ΔAICc > 10. The table shows only fixed-effect terms, but all models also include two random-effect terms: student and the instructor whose classes’ students were enrolled in.
bTime = exam number in a course; Cum.GPA = cumulative college GPA at start of introductory biology series; Gender = student’s gender identity; SES = students’ socioeconomic status; W.Diff = median weighted difficulty of an exam; Percent CR = percentage of constructed-response question on an exam; Course = the three courses that are part of the introductory biology sequence.