Objective: Models of addiction often posit bidirectional and dynamic associations between constructs relevant to the etiology and maintenance of addictive behaviors. The cross-lagged panel model (CLPM) is commonly used in addiction research but has been critiqued for not appropriately adjusting for between-person variance. Alternatives to the CLPMhave been suggested but remain underutilized. The primary purpose of this article is to highlight interpretational limitations of the CLPMand to provide examples of alternative models. Method: We specified CLPM, Random-Intercept CLPM, and a Latent Curve Model with Structured Residuals using four waves of data from Project MATCH (n = 1,201). We modeled prospective relations among depression symptoms and temptation to drink. Substantive inferences and assumptions across models were compared. Results: The CLPMprovided the most evidence of significant cross-lagged paths but the poorest fit to the data compared to other models. Alternative models found little evidence of prospective within-person associations, and more evidence for between-person associations and wave-specific within-person relations between depression symptoms and temptation to drink. Conclusions: This study highlights shortcomings of the CLPM and details alternative models to consider. Addiction researchers should consider alternatives to the CLPMto more optimally delineate relations among constructs across time.
- Cross-lagged panel model
- Latent curve models with structured residuals
- Random-intercept cross-lagged panel model