Skip to main content
. Author manuscript; available in PMC: 2016 Jan 12.
Published in final edited form as: Sci Stud Read. 2016 Jan 5;20(1):20–33. doi: 10.1080/10888438.2015.1107073

Table 2.

Proportion of variance accounted for by the models.

Dependent Variable Sample used Upper-bound for the regression based models Random Forests
Total Reading Time
All .083 .082
Test .075 .080
Skipping
All .029 .027
Test .022 .026
Comprehension Score
All .236 .174
Test .171 .173

The R2 of the full regression model with all predictors is reported as the upper bound for Dominance Analysis and Multimodel Inference. The R2 values for the All sample are based on the models trained on the entire dataset, and are calculated based either on the models’ fitted values to the entire dataset (for full regression model), or on the model's prediction to the out-of-bag samples (for Random Forests). The R2 values for the Test sample are based on the models trained on the 70% random training subsamples and are calculated using the models’ prediction to the remaining 30% testing subsamples, averaged over 100 runs.