Using extreme gradient boosting to estimate community effects on school readiness

The overall purpose of this project is to evaluate features that predict students’ readiness of school upon kindergarten entry. This dashboard represents preliminary results that do not yet include the full data. Results are displayed in separate tabs for each outcome: Early mathematics, Letter Sound identification, and aggregate scores from a measure of assessing children’s self-regulation and interpersonal skills (referred to broadly as Social).

For questions or further information, please contact Daniel Anderson.

Citation

Please refer to this dashboard using the following citation.

Anderson, D., Rosenberg, J. M., Sáez, L., & Seeley, J. R. (September, 2020). Using extreme gradient boosting to estimate community effects on school readiness. Poster presented at the inaugural Conference on Educational Data Science hosted by Stanford University.

Pre-K provider data, including geographic data, licensure, and other quality indicators
- Schools associated with pre-K provider data by computing a weighted (by physical distance from the school) sum of features for all providers within a 40 mile radius
203 total features, including geographical location of school and other enrollment data
District-level covariates included from the Stanford Educational Data Archive

Summary of features

Pre-K Features

3.00

Test RMSE

2.39

Test MAE

0.11

Test

Hyperparameter tuning

Feature interpretation

8.80

Test RMSE

6.76

Test MAE

0.17

Test

Hyperparameter tuning

Feature interpretation