Analytic methods for modeling longitudinal data from rolling therapy groups with membership turnover.

Area of Science:

• Psychology
• Statistics
• Behavioral Science

Background:

• Treatment groups for emotional and behavioral issues often use rolling admission, leading to continuous membership turnover.
• Existing analytic methods for rolling group data are frequently flawed or inadequate.

Purpose of the Study:

• To highlight the analytic challenges of rolling group data.
• To propose latent class pattern mixture models (LCPMMs) as a robust method for analyzing such data.
• To compare LCPMMs with traditional latent growth models (LGMs).

Main Methods:

• Overview of analytic challenges in rolling group data.
• Proposal and illustration of latent class pattern mixture models (LCPMMs).
• Application of LCPMMs to data from an alcoholism treatment trial (N=128).

Main Results:

• Latent class pattern mixture models (LCPMMs) provide a statistically and conceptually sound approach for rolling group data.
• Inferences about treatment efficacy differ between LCPMMs and standard group-clustered latent growth models (LGMs).
• Latent growth models (LGMs) may yield overly liberal results for rolling group data.

Conclusions:

• Latent class pattern mixture models (LCPMMs) are a recommended analytic strategy for rolling group data.
• Further research into data analytic methods for groups with membership turnover is crucial for ecological validity in behavioral treatment research.