Latent Transition Analysis

Latent transition analysis (LTA) and latent class analysis (LCA) are closely related methods. LCA identifies unobservable (latent) subgroups within a population based on individuals’ responses to multiple observed variables. LTA is an extension of LCA that uses longitudinal data to identify movement between the subgroups over time.

A latent transition is movement from one latent subgroup to another over time. We refer to the subgroups as statuses rather than classes to help maintain the distinction between cross-sectional and longitudinal studies. Latent transition analysis (LTA) enables researchers to estimate how membership in the subgroups changes over time. In order to perform LTA, you must have longitudinal data.

Introductory Example: Changes in Teen Sexual Risk

In order to better understand high-risk sexual behavior and its relationship to substance use, several measures of teen dating, sex, and substance use behaviors were selected from rounds 2, 3, and 4 of the National Longitudinal Survey of Youth 1997. This information is drawn from Lanza and Collins (2008) and is also discussed in chapter 8 of Latent Class and Latent Transition Analysis by Linda Collins and Stephanie Lanza.

Sexual/Dating Activity

The researchers fit models with two, three, four, five, and six statuses. The five status model was selected based on fit statistics and interpretability. The researchers labeled the statuses Nondaters, Daters, Monogamous, Multipartner Safe, and Multipartner Exposed. Measurements were taken when participants were 17 or 18 years old (Time 1) and again one and two years later. The most likely responses of members of each status are described below.

  • Nondaters – 0 dating partners, no sex in the past year – 18.6% prevalance at Time 1
  • Daters – 2 or more dating partners, no sex in the past year – 28.9 %
  • Monogamous – 1 dating partner, 1 sex partner – 11.7 %
  • Multipartner Safe – 2 or more dating and sexual partners; used condom every time – 23.1%
  • Multipartner Exposed – 2 or more dating and sexual partners; did not use a condom every time (at risk for STDs) – 17.7%

At time one, when participants in the study were 17 or 18 years old, 17.7% of the sample was in the highest risk latent status, Multipartner Exposed. By time two this status had grown to 20.6% of the population. By time three, nearly 1-in-4 people were in the Multipartner Exposed status. These snapshots of behavior could have been revealed by LCA, but LTA also provides information about movement between statuses over time. Across the study, members of the Multipartner Exposed status were the most stable in their status membership. Somewhat surprisingly, individuals who were most likely to transition into this higher-risk status were members of the Monogamous status, suggesting that the Monogamous status might be an important target for prevention efforts.

The study also examined the association between substance use and sexual risk behavior. Alcohol, cigarette, and marijuana use were all significantly related to sexual risk behavior at time one, although alcohol and marijuana use were stronger predictors of membership in the higher risk Multipartner Exposed status than cigarette use. Past-year drunkenness predicted transitions from the Nondaters and Daters statuses to the Multipartner Exposed status.

Our research on LTA allows scientists to better understand the impact of exposure to multiple risks and the nature of changes in complex behaviors, particularly when change can be represented as movement between discrete categories or stages. This makes it possible to identify groups of at-risk individuals. This analysis was conducted using PROC LTA, a free, easy-to-use SAS procedure for LTA.

References

Lanza, S. T., & Collins, L. M. (2008). A new SAS procedure for latent transition analysis: Transitions in dating and sexual behavior. Developmental Psychology, 42(2), 446-456. PMCID: PMC2846549  View article

Collins, L. M., & Lanza, S. T. (2010). Latent class and latent transition analysis: With applications in the social, behavioral, and health sciences. New York, NY: John Wiley & Sons, Inc.