Exploring father-adolescent closeness: A random forest approach.

Area of Science:

• Family Science
• Computational Social Science
• Artificial Intelligence in Social Research

Background:

• Fatherhood research faces recruitment, retention, and data complexity challenges.
• Machine learning (ML) offers solutions for analyzing large, complex datasets and handling missing data.
• ML can mitigate methodological hurdles in studying father-child relationships.

Purpose of the Study:

• To demonstrate the utility of random forest, a machine learning algorithm, in family science.
• To identify key predictors of father-adolescent closeness.
• To enhance regression models using ML-informed variable selection.

Main Methods:

• Applied random forest algorithm to predict father-adolescent closeness.
• Utilized data from the Future of Families and Child Wellbeing Study (n=2,927).
• Included 131 predictors measured in the first decade of childhood.

Main Results:

• Fathers' residential status with the child was the strongest predictor of closeness.
• Random forest improved variable selection for enhanced regression models.
• Demonstrated ML's capability in identifying complex family context predictors.

Conclusions:

• Random forest is a valuable tool for family scientists studying complex relationships.
• Incorporating artificial intelligence, like random forest, advances family science research.
• This approach offers new directions for understanding father-child dynamics.