Drawing causal inferences using propensity scores: a practical guide for community psychologists

Area of Science:

• Community Psychology
• Developmental Psychology
• Biostatistics

Background:

• Observational data often contains confounding variables, hindering causal inference in community psychology.
• Propensity score methods offer a structured approach to address confounding in observational studies.

Purpose of the Study:

• To demonstrate propensity score methods for causal inference in community psychology research.
• To examine the causal effect of preschool programs (Head Start vs. parental care) on reading development.

Main Methods:

• The study applied three propensity score techniques: weighting, matching, and subclassification.
• These methods were used to analyze observational data on children's preschool experiences and kindergarten reading scores.

Main Results:

• Unadjusted analysis suggested parental care led to higher reading scores.
• Propensity score adjustments reduced the estimated causal effect by over 50%.
• No evidence indicated improved reading outcomes for children who received parental care instead of Head Start.

Conclusions:

• Propensity score methods effectively mitigate confounding in observational studies.
• Preschool experiences like Head Start do not show a detrimental effect on reading development compared to parental care.
• The study provides practical tools (SAS/R syntax) for community psychologists to conduct causal inference.