Identification and Sensitivity Analysis for Average Causal Mediation Effects with Time-Varying Treatments and Mediators: Investigating the Underlying Mechanisms of Kindergarten Retention Policy

Considering that causal mechanisms unfold over time, it is important to investigate the mechanisms over time, taking into account the time-varying features of treatments and mediators. However, identification of the average causal mediation effect in the presence of time-varying treatments and mediators is often complicated by time-varying confounding. This article aims to provide a novel approach to uncovering causal mechanisms in time-varying treatments and mediators in the presence of time-varying confounding. We provide different strategies for identification and sensitivity analysis under homogeneous and heterogeneous effects. Homogeneous effects are those in which each individual experiences the same effect, and heterogeneous effects are those in which the effects vary over individuals. Most importantly, we provide an alternative definition of average causal mediation effects that evaluates a partial mediation effect; the effect that is mediated by paths other than through an intermediate confounding variable. We argue that this alternative definition allows us to better assess at least a part of the mediated effect and provides meaningful and unique interpretations. A case study using ECLS-K data that evaluates kindergarten retention policy is offered to illustrate our proposed approach.     

纵向数据的中介效应分析

目前中介效应检验主要是基于截面数据,但许多时候截面数据的中介分析不适合进行因果推断,因而需要收集历时性的纵向数据,进行纵向数据的中介分析。评介了基于交叉滞后面板模型、多层线性模型和潜变量增长模型的纵向数据的中介分析方法及其四个发展。第一,中介效应随时间变化,如连续时间模型、多层时变系数模型。第二,中介效应随个体变化,如随机效应的交叉滞后面板模型和多层自回归模型。第三,中介模型的整合,如交叉滞后面板模型与多层线性模型整合为多层自回归模型。第四,中介检验方法的发展,建议使用Monte Carlo、Bootstrap和贝叶斯法进行纵向数据的中介分析。总结出一个纵向数据的中介分析流程并给出相应的Mplus程序。随后展望了纵向数据的中介分析的拓展方向。     

Demographic variables, smoking variables, and outcome across five studies.

OBJECTIVE: Intervention effectiveness can potentially be affected by membership in different demographic subgroups (race, ethnicity, gender, age, and education level) or smoking behavior variables (time to first cigarette, longest previous quit attempt, number of attempts in the past year, number of cigarettes, and stage of change). Previous research on these 2 sets of variables has produced mixed results. DESIGN: This secondary data analysis combined data from 5 effectiveness trials (a random-digit-dial sample [N=1,358], members of an HMO [N=207], parents of students recruited for a school-based study [N=347], patients from an insurance provider list [N=535], and employees [N=175]) in which smokers were all proactively recruited from a defined population and all received the same expert system intervention. The intervention produced a consistent 22% to 26% point prevalence cessation rate across the 5 studies. MAIN OUTCOME MEASURES: The main outcome measures were 24-hr point prevalence, 7-day point prevalence, 30-day prolonged abstinence, and 6-month prolonged abstinence. RESULTS: There were no significant differences in outcome across gender, race, and ethnicity subgroups. There were significant differences and small effect sizes for age and education subgroups. There were significant differences and large effect sizes for all 5 smoking behavior variables. DISCUSSION: Demographic variables are static variables, whereas the smoking variables are more dynamic, that is, open to change. Given the dynamic nature of the smoking variables and the large effect sizes, interventions tailored on the smoking variables should be more successful.     

Patterns of self-initiated smoking cessation among young adults

Prochaska and DiClemente's (1984) cyclic-stage model of self-initiated smoking cessation divides the cessation process into five stages. This model was applied to a young adult population to determine the cross-sectional distribution of stages and the frequency and pattern of changes among stages over time. Compared to older adults, the distribution of the stages differed substantially: There were twice as many relapsers and only half as many maintainers among young adults. One-year changes in stages were examined using a static model, which did not take into account the cyclic nature of the change process, and a more realistic dynamic model, which did. Both models, especially the dynamic model, suggested substantially more movement among stages in younger than in older adults.     

Dynamic Factor Analysis Models With Time-Varying Parameters

Dynamic factor analysis models with time-varying parameters offer a valuable tool for evaluating multivariate time series data with time-varying dynamics and/or measurement properties. We use the Dynamic Model of Activation proposed by Zautra and colleagues (Zautra, Potter, & Reich, 1997) as a motivating example to construct a dynamic factor model with vector autoregressive relations and time-varying cross-regression parameters at the factor level. Using techniques drawn from the state-space literature, the model was fitted to a set of daily affect data (over 71 days) from 10 participants who had been diagnosed with Parkinson's disease. Our empirical results lend partial support and some potential refinement to the Dynamic Model of Activation with regard to how the time dependencies between positive and negative affects change over time. A simulation study is conducted to examine the performance of the proposed techniques when (a) changes in the time-varying parameters are represented using the true model of change, (b) supposedly time-invariant parameters are represented as time-varying, and (c) the time-varying parameters show discrete shifts that are approximated using an autoregressive model of differences.     

A joint modeling framework of responses and response times to assess learning outcomes

Abstract

A general modeling framework of response accuracy and response times is proposed to track skill acquisition and provide additional diagnostic information on the change of latent speed in a learning environment. This framework consists of two types of models: a dynamic response model that captures the response accuracy and the change of discrete latent attribute profile upon factors such as practice, intervention effects, and other latent and observable covariates, and a dynamic response time model that describes the change of the continuous response latency due to change of latent attribute profile. These two types of models are connected through a parameter, describing the change rate of the latent speed through the learning process, and a covariate defined as a function of the latent attribute profile. A Bayesian estimation procedure is developed to calibrate the model parameters and measure the latent variables. The estimation algorithm is evaluated through several simulation studies under various conditions. The proposed models are applied to a real data set collected through a spatial rotation diagnostic assessment paired with learning tools.     

老年人认知功能的变化轨迹：基于潜变量增长模型的分析

摘 要 本研究采用中国老年健康影响因素跟踪调查（CLHLS）的4次数据（2002,2005,2008,2011）,对老年人认知功能的变化趋势以及影响因素进行了探讨。结果显示：（1）老年人认知功能在4次测查中呈非线性的下降趋势。（2）日常生活活动能力较低的个体其认知功能也较低;老年人读书年限越高,其认知功能水平越高;女性的认知功能水平低于男性;不饮酒的老年人认知功能低于饮酒的老年人。（3）读书年限与饮酒会正向预测模型的斜率。     

Latent growth curve analyses of peer and parent influences on smoking progression among early adolescents.

Social influences on smoking uptake were examined in latent growth curve analyses of data from 1,320 youths assessed 5 times during 6th to 9th grade. Initial smoking stage predicted increases in number of friends who smoked, indicating selection; however, initial number of friends who smoked did not predict smoking stage progression, indicating no significant effect of socialization. Associations over time among smoking stage progression, affiliation with friends who smoke, and parenting behaviors were significant, suggesting dynamic, reciprocal relationships. Parental involvement, monitoring, and expectations provided direct protective effects against smoking progression as well as indirect effects, by limiting increases in number of friends who smoke. These results are consistent with the peer selection hypothesis, confirm the powerful association over time of social influences with smoking, and provide the first evidence that parenting behavior may protect against smoking progression by limiting increases in number of friends who smoke.     

The interplay of received social support and self-regulatory factors in smoking cessation

Objective: In smoking cessation, individual self-regulation and social support have both proven to be useful. However, the roles of self-regulatory processes and social support are mostly examined separately. The present study aims at examining the unique and joint interactive effects of self-regulation as specified in the health action process approach (HAPA) and social support on smoking cessation. The study tested whether social support can compensate for low levels of self-regulation or whether synergistic effects emerge.

Design & Measures: Around a self-set quit date, 99 smokers completed baseline questionnaires on HAPA-variables, smoking-specific received social support and smoking cessation (continuous abstinence and point prevalence), with a follow-up Cpproximately 29?days after the quitdate.

Results: Social support moderated the association between volitional self-efficacy and smoking, as well as coping planning and smoking but not between action planning and smoking. No compensatory effect of social support for lower levels of individual regulation emerged but the combination of high levels of the individual variables and social support was related to successful smoking cessation, indicating a synergistic effect.

Conclusions: The results confirm the importance of examining both self-regulation and social factors in smoking cessation. This should be considered when developing future interventions for smoking cessation.     

Extreme Response Style and the Measurement of Intra-Individual Variability in Affect

Extreme response style (ERS) has the potential to bias the measurement of intra-individual variability in psychological constructs. This paper explores such bias through a multilevel extension of a latent trait model for modeling response styles applied to repeated measures rating scale data. Modeling responses to multi-item scales of positive and negative affect collected from smokers at clinic visits following a smoking cessation attempt revealed considerable ERS bias in the intra-individual sum score variances. In addition, simulation studies suggest the magnitude and direction of bias due to ERS is heavily dependent on the mean affect level, supporting a model-based approach to the study and control of ERS effects. Application of the proposed model-based adjustment is found to improve intra-individual variability as a predictor of smoking cessation.     

基于结构方程模型的有调节的中介效应分析

有调节的中介模型是中介过程受到调节变量影响的模型。指出了目前有调节的中介效应分析普遍存在的问题：当前有调节的中介效应检验大多使用多元线性回归分析,忽略了测量误差;而基于结构方程模型(SEM)的有调节的中介效应分析需要产生乘积指标,又会面临乘积指标生成和乘积项非正态分布的问题。在简介潜调节结构方程(LMS)方法后,建议使用LMS方法得到偏差校正的bootstrap置信区间来进行基于SEM的有调节的中介效应分析。总结出一个有调节的中介SEM分析流程,并有示例和相应的Mplus程序。文末展望了LMS和有调节的中介模型的发展方向。     

Cross-Sectional Analysis of Longitudinal Mediation Processes

Statistical mediation analysis can help to identify and explain the mechanisms behind psychological processes. Examining a set of variables for mediation effects is a ubiquitous process in the social sciences literature; however, despite evidence suggesting that cross-sectional data can misrepresent the mediation of longitudinal processes, cross-sectional analyses continue to be used in this manner. Alternative longitudinal mediation models, including those rooted in a structural equation modeling framework (cross-lagged panel, latent growth curve, and latent difference score models) are currently available and may provide a better representation of mediation processes for longitudinal data. The purpose of this paper is twofold: first, we provide a comparison of cross-sectional and longitudinal mediation models; second, we advocate using models to evaluate mediation effects that capture the temporal sequence of the process under study. Two separate empirical examples are presented to illustrate differences in the conclusions drawn from cross-sectional and longitudinal mediation analyses. Findings from these examples yielded substantial differences in interpretations between the cross-sectional and longitudinal mediation models considered here. Based on these observations, researchers should use caution when attempting to use cross-sectional data in place of longitudinal data for mediation analyses.     

新世纪20年国内心理统计方法研究回顾

新世纪头20年, 国内心理学11本专业期刊一共发表了213篇统计方法研究论文。研究范围主要包括以下10类(按论文篇数排序)：结构方程模型、测验信度、中介效应、效应量与检验力、纵向研究、调节效应、探索性因子分析、潜在类别模型、共同方法偏差和多层线性模型。对各类做了简单的回顾与梳理。结果发现, 国内心理统计方法研究的广度和深度都不断增加, 研究热点在相互融合中共同发展; 但综述类论文比例较大, 原创性研究论文比例有待提高, 研究力量也有待加强。     

What is the right thing at the right time? Interactions between stages and processes of change among smokers who make a quit attempt.

C. A. Perz, C. C. DiClemente, and J. P. Carbonari (1996) claim support for the transtheoretical model notion that success in smoking cessation involves doing the right thing at the right time: emphasising experiential change processes during the contemplation and preparation stages and shifting to behavioral process activities during action. A key methodological limitation of Perz et al. was their failure to control for stage of change, a measure that has been shown to be predictive of cessation. This study replicates the prospective findings of Perz et al. in a different data set, then controls for stage of change when it is predictive of cessation, and finds that the measures of "appropriate" change process use developed by Perz et al. no longer predict cessation. The authors conclude that stage of change, in particular the distinction between smoking and not smoking, is more important than change process use in predicting cessation outcomes.     

Dynamic effects of self-efficacy on smoking lapse and relapse.

Self-efficacy (SE) is thought to be critical to success in smoking cessation both as an individual difference and as a dynamic process after a quit attempt. In this study, 214 smokers used palm-top computers to record day-to-day variations in SE during 4 weeks after quitting. SE remained at high and stable levels prior to a 1st lapse but decreased and became more variable thereafter. The authors used event history models with time-varying covariates to assess the effect of daily SE on lapse and relapse risk. Daily SE measures predicted an initial lapse on the subsequent day. However, this relationship was accounted for by stable baseline differences in SE (assessed by questionnaire), rather than by day-to-day dynamics in SE. Progression from 1st lapse to relapse was also examined. In this instance, daily SE predicted subsequent relapse risk, even when baseline SE and concurrent smoking were accounted for, suggesting the importance of SE dynamics for this stage of the relapse process.     

A Life-Course View on Ageing Consumers: Old-Age Trajectories and Gender Differences

This study presents a dynamic, model-based view of consumers’ ageing developments, focused on gender differences, to uncover the pathways and socioeconomic transitions that female and male consumers take through old age. The analysis of longitudinal survey data spanning 15 years uses a latent Markov dynamic cluster model with transitions over time. The resulting life courses allow an exploration of lifestyle-related changes in multiple consumer well-being variables beyond age 50. Substantial well-being differences appear in the ageing paths of men and women. In both cases, a dominant chronological sequence through old age is complemented by less common transitions, rarely associated with advanced age. Although the model does not use chronological age as an independent variable, it outperforms purely agebased, or age- cohort-, and period-based models in predicting old-age consumer wellbeing. These results highlight the importance of considering within-cohort diversity when modelling the accompaniments of old age: while some older consumers enjoy active lifestyles, others of similar age succumb to depression and loneliness, rendering age an insufficient predictor of well-being states. In the future, the presented model could be matched with other, even cross-sectional, consumer survey data to help predict various dynamics in the ageing consumer population.

     

Rater Model Using Signal Detection Theory for Latent Differential Rater Functioning

Differential rater functioning (DRF) occurs when raters show evidence of exercising differential severity or leniency when scoring examinees within different subgroups. Previous studies of DRF have examined rater bias using manifest variables (e.g., use of covariates) to determine the subgroups. These manifest variables include gender and the ethnicity of the examinee. For example, a rater may score males more severely. Ideally, each rater’s severity should be invariant across subgroups. This study examines DRF in the context of latent subgroups that classify possible sources of DRF based on raters’ scoring behavior rather than manifest factors. An extension of the latent class signal detection theory (LC-SDT) model for identifying DRF is proposed and examined using real-world data and simulations. Results from real-world data show that the signal detection approach leads to an effective method to identify latent DRF. Simulations with varying sample sizes and conditions of rater precision were shown to recover parameters at an adequate level, supporting its use to identify latent DRF in large-scale data. These findings suggest that the DRF extension of the LC-SDT can be a useful model to examine characteristics of raters and add information that can aid rater training.     

Finite Mixtures of Hidden Markov Models for Longitudinal Responses Subject to Drop out

Abstract

Drop out is a typical issue in longitudinal studies. When the missingness is non-ignorable, inference based on the observed data only may be biased. This paper is motivated by the Leiden 85+ study, a longitudinal study conducted to analyze the dynamics of cognitive functioning in the elderly. We account for dependence between longitudinal responses from the same subject using time-varying random effects associated with a heterogeneous hidden Markov chain. As several participants in the study drop out prematurely, we introduce a further random effect model to describe the missing data mechanism. The potential dependence between the random effects in the two equations (and, therefore, between the two processes) is introduced through a joint distribution specified via a latent structure approach. The application of the proposal to data from the Leiden 85+ study shows its effectiveness in modeling heterogeneous longitudinal patterns, possibly influenced by the missing data process. Results from a sensitivity analysis show the robustness of the estimates with respect to misspecification of the missing data mechanism. A simulation study provides evidence for the reliability of the inferential conclusions drawn from the analysis of the Leiden 85+ data.     

A time-varying effect model for intensive longitudinal data

Understanding temporal change in human behavior and psychological processes is a central issue in the behavioral sciences. With technological advances, intensive longitudinal data (ILD) are increasingly generated by studies of human behavior that repeatedly administer assessments over time. ILD offer unique opportunities to describe temporal behavioral changes in detail and identify related environmental and psychosocial antecedents and consequences. Traditional analytical approaches impose strong parametric assumptions about the nature of change in the relationship between time-varying covariates and outcomes of interest. This article introduces time-varying effect models (TVEMs) that explicitly model changes in the association between ILD covariates and ILD outcomes over time in a flexible manner. In this article, we describe unique research questions that the TVEM addresses, outline the model-estimation procedure, share a SAS macro for implementing the model, demonstrate model utility with a simulated example, and illustrate model applications in ILD collected as part of a smoking-cessation study to explore the relationship between smoking urges and self-efficacy during the course of the pre- and postcessation period.