Confidence interval estimation for standardized effect sizes in multilevel and latent growth modeling

Area of Science:

• Statistics
• Biostatistics
• Psychometrics

Background:

• Multilevel and latent growth models are common in clinical trials.
• These models assess group differences in outcome trajectories.
• The treatment effect can be standardized as Cohen's d.

Purpose of the Study:

• To present methods for calculating confidence intervals (CIs) for the treatment effect size.
• To provide equations for estimating CIs in multilevel models.

Main Methods:

• Derived two sets of equations for CI estimation.
• Illustrated usage with data from the National Youth Study.
• Validated CIs using a Monte Carlo simulation study.

Main Results:

• Demonstrated the equivalence of the two CI estimation methods.
• Found minimal bias in the new CI for effect size compared to existing methods.
• Confirmed reliability across varying effect sizes and sample sizes.

Conclusions:

• Provides accessible equations for estimating CIs for a popular effect size.
• Facilitates accurate interpretation of treatment effects in controlled clinical trials.