Do Different Methods for Modeling Age-Graded Trajectories Yield Consistent and Valid Results?

Area of Science:

• Developmental Psychology
• Longitudinal Data Analysis
• Quantitative Methods

Background:

• Age-sequenced individual attribute data are increasingly utilized across diverse research fields.
• A lack of consensus exists regarding the optimal methods for identifying discrete trajectories and assigning individuals to them.

Purpose of the Study:

• To compare the performance of various statistical methods in identifying the number of trajectories and assigning individuals to trajectory groups.
• To evaluate the consistency and accuracy of different trajectory modeling approaches using both real and simulated data.

Main Methods:

• Modeled trajectory data using "naïve" methods, optimal matching, grade of membership models, and three types of finite mixture models.
• Utilized simulated data with known trajectory parameters to assess method accuracy against ground truth.

Main Results:

• Significant discrepancies were observed among the tested methods regarding the inferred number of trajectories.
• Method-based inferences frequently deviated from the true number of trajectories in simulated data.
• Individual assignments to trajectory groups also varied considerably across the different analytical approaches.

Conclusions:

• Researchers must exercise caution when interpreting findings derived from trajectory analysis methods.
• Avoid over-reliance on point estimates for the number of trajectories and treat individual group assignments as provisional.