Peer-reviewed article
Using LASSO to Assist Imputation and Predict Child Well-being
Diana M. Stanescu, Erik Wang, Soichiro Yamauchi ·
Summary
Written by AI, not yet AICS's view · claude-sonnet-5-5This article describes a method for predicting children's well-being using data from the Fragile Families and Child Wellbeing Study, which represents births in large U.S. cities. The authors used a statistical technique called LASSO to prepare the data, filled in missing values with the Amelia algorithm, and then used LASSO again to make predictions. They report results for six outcome variables. The method worked best for material hardship, where its out-of-sample mean squared error of 0.019 was the lowest among all submissions in the Fragile Families Challenge.
Key points
- Data came from the Fragile Families and Child Wellbeing Study, representative of births in large U.S. cities.
- The approach combined LASSO preprocessing, Amelia imputation of missing data, and LASSO prediction.
- Performance was reported for six outcome variables, with the best result for material hardship.
- Out-of-sample mean squared error for material hardship was 0.019, the lowest among all Fragile Families Challenge submissions.
- Among variables with high predictive power, mother-survey variables dominated, and past components of material hardship strongly predicted current material hardship.
Limitations
The abstract does not give results for the five other outcome variables or details of how the models were validated. It describes a prediction method rather than an intervention, so it cannot show what causes child well-being outcomes or what would improve them.
Abstract
This article documents an approach to predicting children’s well-being using data from the Fragile Families and Child Wellbeing Study, which are representative of births in large U.S. cities. The authors use the least absolute shrinkage and selection operator (LASSO) to preprocess the data. They then apply the Amelia algorithm to impute missing data. Finally, they use LASSO again for prediction with the imputed data. The authors report the performance of this approach for six outcome variables. The approach achieves the best performance for the variable material hardship. The out-of-sample mean squared error of the authors’ prediction is 0.019, the lowest among all submissions in the Fragile Families Challenge. The authors find that among variables with high predictive power, variables from mother surveys dominate. Furthermore, components of material hardship in the past strongly predict current material hardship.