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

Abstract

Data on age-sequenced trajectories of individuals’ attributes are used for a growing number of research purposes. However, there is no consensus about which method to use to identify the number of discrete trajectories in a population or to assign individuals to a specific trajectory group. We modeled real and simulated trajectory data using “naïve” methods, optimal matching, grade of membership models, and three types of finite mixture models. We found that these methods produced inferences about the number of trajectories that frequently differ (1) from one another and (2) from the truth as represented by simulation parameters. We also found that they differed in the assignment of individuals to trajectory groups. In light of these findings, we argue that researchers should interpret results based on these methods cautiously, neither reifying point estimates about the number of trajectories nor treating individuals’ trajectory group assignments as certain.

Trajectories of ethnic neighbourhood change: Spatial patterns of increasing ethnic diversity

“…Although we cannot be certain about the “true” number of latent trajectories, descriptive statistics (see Table 3) and geographical maps (see Figures 2 and 3) of our five classes correspond to the known ethnic distribution in Dutch cities. The uncertainty around the true number of latent trajectories is especially problematic when trajectories are used as dependent or independent variables in subsequent analyses (Warren et al, ). The goal of the present study is however mainly descriptive, and although we cannot be certain about the true number of trajectories, four‐ and six‐class models showed similar trajectories over time.…”

The Fragmented Evolution of Racial Integration since the Civil Rights Movement

“…Problems introduced by the transformation aside, some evidence suggests that even when using the correct specification, growth mixture models can underestimate the correct number of classes (Warren et al 2015). Our robustness check mitigates some concerns related to correctly identifying the true number of latent trajectories.…”

“…Robustness to the Number of Trajectories Warren et al (2015) warn of overstating the certainty with which one asserts to have identified the correct number of classes from growth mixture models. We determined that the model contained 11 trajectories based on three factors: a substantial decline in the rate of BIC improvement, a statistically significant LMR-LRT value, and a high entropy score.…”

Trajectories of Union Transition in Emerging Adulthood: Socioeconomic Status and Race/Ethnicity Differences in the National Longitudinal Survey of Youth 1997 Cohort

“…As with nonparametric models that suffer from the “curse of dimensionality,” researchers should consider these tradeoffs when deciding whether to use multichannel sequence analysis. Finally, we acknowledge criticism on the use of sequence analysis (and other similar methods) in revealing “true” underlying trajectory types (Warren, Luo, Halpern‐Manners, Raymo, & Palloni, ). Warren et al () found that different methods for modeling age‐graded trajectories may yield slightly different results.…”

“…Finally, we acknowledge criticism on the use of sequence analysis (and other similar methods) in revealing “true” underlying trajectory types (Warren, Luo, Halpern‐Manners, Raymo, & Palloni, ). Warren et al () found that different methods for modeling age‐graded trajectories may yield slightly different results. As highlighted previously, however, some trajectory models may not be easily estimable with high‐resolution data.…”