Xianming Tan

Latent Class Analysis With Distal Outcomes: A Flexible Model-Based Approach

Although prediction of class membership from observed variables in latent class analysis is well understood, predicting an observed distal outcome from latent class membership is more complicated. A flexible model-based approach is proposed to empirically derive and summarize the class-dependent density functions of distal outcomes with categorical, continuous, or count distributions. A Monte Carlo simulation study is conducted to compare the performance of the new technique to 2 commonly used classify-analyze techniques: maximum-probability assignment and multiple pseudoclass draws. Simulation results show that the model-based approach produces substantially less biased estimates of the effect compared to either classify-analyze technique, particularly when the association between the latent class variable and the distal outcome is strong. In addition, we show that only the model-based approach is consistent. The approach is demonstrated empirically: latent classes of adolescent depression are used to predict smoking, grades, and delinquency. SAS syntax for implementing this approach using PROC LCA and a corresponding macro are provided.

Effect Size, Statistical Power and Sample Size Requirements for the Bootstrap Likelihood Ratio Test in Latent Class Analysis.

Selecting the number of different classes that will be assumed to exist in the population is an important step in latent class analysis (LCA). The bootstrap likelihood ratio test (BLRT) provides a data-driven way to evaluate the relative adequacy of a (K – 1)-class model compared to a K-class model. However, very little is known about how to predict the power or the required sample size for the BLRT in LCA. Based on extensive Monte Carlo simulations, we provide practical effect size measures and power curves that can be used to predict power for the BLRT in LCA given a proposed sample size and a set of hypothesized population parameters. Estimated power curves and tables provide guidance for researchers wishing to size a study to have sufficient power to detect hypothesized underlying latent classes.

Eliminating Bias in Classify-Analyze Approaches for Latent Class Analysis

Despite recent methodological advances in latent class analysis (LCA) and a rapid increase in its application in behavioral research, complex research questions that include latent class variables often must be addressed by classifying individuals into latent classes and treating class membership as known in a subsequent analysis. Traditional approaches to classifying individuals based on posterior probabilities are known to produce attenuated estimates in the analytic model. We propose the use of a more inclusive LCA to generate posterior probabilities; this LCA includes additional variables present in the analytic model. A motivating empirical demonstration is presented, followed by a simulation study to assess the performance of the proposed strategy. Results show that with sufficient measurement quality or sample size, the proposed strategy reduces or eliminates bias.

Statistical models for longitudinal zero‐inflated count data with applications to the substance abuse field

This study fills in the current knowledge gaps in statistical analysis of longitudinal zero-inflated count data by providing a comprehensive review and comparison of the hurdle and zero-inflated Poisson models in terms of the conceptual framework, computational advantage, and performance under different real data situations. The design of simulations represents the special features of a well-known longitudinal study of alcoholism so that the results can be generalizable to the substance abuse field. When the hurdle model is more natural under the conceptual framework of the data, the zero-inflated Poisson model tends to produce inaccurate estimates. Model performance improves with larger sample sizes, lower proportions of missing data, and lower correlations between covariates. The simulation also shows that the computational strength of the hurdle model disappears when random effects are included.

A Social Network–Informed Latent Class Analysis of Patterns of Substance Use, Sexual Behavior, and Mental Health: Social Network Study III, Winnipeg, Manitoba, Canada

OBJECTIVES We assessed whether a meaningful set of latent risk profiles could be identified in an inner-city population through individual and network characteristics of substance use, sexual behaviors, and mental health status. METHODS Data came from 600 participants in Social Network Study III, conducted in 2009 in Winnipeg, Manitoba, Canada. We used latent class analysis (LCA) to identify risk profiles and, with covariates, to identify predictors of class. RESULTS A 4-class model of risk profiles fit the data best: (1) solitary users reported polydrug use at the individual level, but low probabilities of substance use or concurrent sexual partners with network members; (2) social-all-substance users reported polydrug use at the individual and network levels; (3) social-noninjection drug users reported less likelihood of injection drug and solvent use; (4) low-risk users reported low probabilities across substances. Unstable housing, preadolescent substance use, age, and hepatitis C status predicted risk profiles. CONCLUSIONS Incorporation of social network variables into LCA can distinguish important subgroups with varying patterns of risk behaviors that can lead to sexually transmitted and bloodborne infections.

Nicotine-dependence-varying effects of smoking events on momentary mood changes among adolescents

INTRODUCTION Theories of nicotine addiction emphasize the initial role of positive reinforcement in the development of regular smoking behavior, and the role of negative reinforcement at later stages. These theories are tested here by examining the effects of amount smoked per smoking event on smoking-related mood changes, and how nicotine dependence (ND) moderates this effect. The current study examines these questions within a sample of light adolescent smokers drawn from the metropolitan Chicago area (N=151, 55.6% female, mean 17.7years). INSTRUMENTS Ecological momentary assessment data were collected via handheld computers, and additional variables were drawn from a traditional questionnaire. METHODS Effects of the amount smoked per event on changes in positive affect (PA) and negative affect (NA) after vs. before smoking were examined, while controlling for subject-averaged amount smoked, age, gender, and day of week. ND-varying effects were examined using varying effect models to elucidate their change across levels of ND. RESULTS The effect of the amount smoked per event was significantly associated with an increase in PA among adolescents with low-to-moderate levels of ND, and was not significant at high ND. Conversely, the effect of the amount smoked was significantly associated with a decrease in NA only for adolescents with low levels of ND. CONCLUSIONS These findings support the role of positive reinforcement in early stages of dependent smoking, but do not support the role of negative reinforcement beyond early stages of smoking. Other potential contributing factors to the relationship between smoking behavior and PA/NA change are discussed.

P2-S5.05 Risk Profiles of Winnipeg Street Populations: A latent class analysis

Background Sexually transmitted and bloodborne infection (STBBI) risk is multifaceted and can involve a complex interplay between sexual behaviours, substance abuse and mental health conditions. In Winnipeg, Manitoba Canada we conducted a study to better understand the interconnectedness and overlap of these conditions and behaviours. Methods Data from the Social Network study phase III (SNS III) were collected in the fall and winter of 2009 using semi-structured in-person interviews (n=600). Sampling was by respondent driven sampling and targeted street populations. The average mean age was 37 (SD=14.8) and the gender distribution was relatively equal (males constituted 53%). Latent class analysis was used to identify unobserved or latent subgroups (ie, risk profiles) to explore the extent of overlap between risky sexual behaviours, substance use choice (crack, alcohol, solvents, injection drug use), and mental health conditions. Six individual level items constituting risky behaviours and five network or environmental level risk behaviours were used in the latent class analysis. Individual items included: Ever diagnosed with a mental health condition, ever used crack, daily binge drinking, ever used solvents, ever injected drugs and knowing your sex partner has multiple other sex partners while social network items included: the proportion of your social network members who drink alcohol, use crack, sniff solvents, inject drugs, or are sex partners. Fit indices of G2, AIC, and BIC were used in assessing model fit. Additionally, the model fit was assessed by examining the relationship between items and their conditional latent class by strength of homogeneity (closeness to 0 or 1) and by whether there was evidence of good separation of latent classes. Results The 2-, 3-, 4-, and 5-class LCA models were compared. Goodness of fit indices favoured the 4-class model. For the 4-class model indices were: G2=1115, df=2000, AIC=1209, BIC=1415. Class prevalence of the 4 latent classes were: 31% were at high risk for all individual and network items, 25% constituted another latent class labelled as low-risk, 21% constituted a subgroup who were labelled as “loners” and exhibited high risk for mental health issues as well as individual crack, solvent use, and injection use but no network level correlates while the fourth latent class (23%) was distinguished for engaging in risky sexual behaviours and having these risky behaviours be supported at the social network level. Conclusions Latent class analysis demonstrated that there are indeed subgroups of vulnerable populations who warrant targeted interventions given their different risk profiles. This type of investigation offers a public health population segmentation strategy to plan for future targeted prevention efforts that can more effectively address the special needs of these subgroups of vulnerable populations. Abstract P2-S5.05 Table 1. 4 Latent class model of Winnipeg street population risk profiles Class membership probabilities: Gamma estimates (SEs) Class 1 2 3 4 0.2474 (0.0254) 0.3099 (0.0315) 0.2137 (0.0243) 0.2289 (0.0325) Item response probabilities: ρestimates (SEs) HSA14_: 0.0335 (0.0482) 0.7211 (0.0372) 0.6719 (0.0473) 0.7207 (0.0467) Mental health (even been dx) ALC2_: 0.3754 (0.0461) 0.8494 (0.0627) 0.4599 (0.0505) 0.0623 (0.0433) Binge drinking (>5 drinks) CRS1_: 0.3158 (0.0540) 0.9773 (0.0166) 0.9128 (0.0342) 0.9258 (0.0315) Crack use SSU1_: 0.1425 (0.0349) 0.5256 (0.0414) 0.5068 (0.0525) 0.5541 (0.0516) Solvent use idu1_: 0.0488 (0.0285) 0.6211 (0.0404) 0.6698 (0.0568) 0.6879 (0.0525) Injection drug use SNSX7_new_: 0.1944 (0.0403) 0.7745 (0.0368) 0.0052 (0.0069) 0.6961 (0.0511) Sex partners have multiple part SNID1_: 0.1784 (0.0398) 0.5803 (0.0407) 0.3842 (0.0483) 0.7888 (0.0477) Network members’ IDU SNALC_: 0.6019 (0.0459) 0.9870 (0.0158) 0.4698 (0.0500) 0.4151 (0.0713) Network members’ alcohol use SNCRACK_: 0.0136 (0.0137) 0.6265 (0.0408) 0.3629 (0.0498) 0.4515 (0.0549) Network members‘ crack use SNODU_: 0.4989 (0.0462) 0.8380 (0.0331) 0.4820 (0.0497) 0.5634 (0.0518) Network members’ other drug use SNSX1_: 0.5883 (0.0496) 0.9872 (0.0121) 0.1129 (0.0572) 0.9958 (0.0068) Network members are sex part.