Controlling for cohort effects in accelerated longitudinal designs using continuous- and discrete-time dynamic...

Area of Science:

• Developmental Psychology
• Quantitative Psychology
• Longitudinal Research Methods

Background:

• Accelerated longitudinal designs (ALDs) enable extended developmental inquiry beyond study duration.
• A core assumption is cohort homogeneity, which, if violated, can bias results.
• Systematic examination of cohort effect consequences in ALDs is lacking.

Purpose of the Study:

• To propose and evaluate a method for detecting and controlling cohort differences in ALDs.
• To utilize latent change score models in both discrete and continuous time for this purpose.

Main Methods:

• Employed latent change score (LCS) models in discrete and continuous time.
• Conducted a Monte Carlo simulation study to assess method effectiveness.
• Evaluated various sampling schedules, including measurement frequency and duration.

Main Results:

• LCS models effectively estimated cohort effects across a range of sizes.
• Continuous-time LCS models showed slightly superior performance.
• Cohort effects on asymptotic levels (d) introduced more bias than on initial levels (d₀).
• Models including cohort effects yielded unbiased estimates; omitting them was tenable only for small effects (d₀ ≤ 1, d ≤ .2).
• Study designs with ≥3 measurements over ≥4 years demonstrated optimal performance.

Conclusions:

• The proposed LCS approach reliably detects and controls for cohort differences in ALDs.
• Recommendations are provided for optimizing study design and data analysis in ALDs.
• Accurate developmental trajectory estimation requires accounting for cohort heterogeneity.