Category: Research Methods and Statistics

Symposium

Longitudinal Structural Equation Modeling: An Example to Understand the Mechanisms Underlying Depression Predicting Anxiety

Sunday, November 19
8:30 AM - 10:00 AM
Location: Aqua 300 A & B, Level 3, Aqua Level

Keywords: Research Methods | Statistics | Longitudinal
Presentation Type: Symposium

Longitudinal research presents as a powerful approach to generate a wealth of nuanced literature. However, the processes of longitudinal change are not well understood, and numerous current methods are not particularly adapted to assess fundamental questions concerning mechanisms of change. Absence of knowledge of longitudinal change processes hinders the deepening our understanding of the factors and mechanisms underlying the effectiveness of cognitive behavioral therapies. In this presentation, we assess the processes of longitudinal change using structural equation modeling (SEM) mediation analyses, by focusing on the direct and indirect effects in construing mechanisms of change. We provide an illustrative example which examines depression predicting generalized anxiety (GA) measured 18 years apart. Perceived controllability (PC) and personal (PM) were examined as theoretical mediators in the longitudinal relationship. 4,963 community-dwelling middle-aged adults (M = 46.46 ± 12.51 years, 53% female, 85% college educated) participated in three waves of data collection over 18 years, with each wave approximately nine years apart from another. The Sense of Control scale (Lachman & Weaver, 1998) was used to assess PC and PM. We controlled for baseline GA in every step of the analyses The overall pattern of practical fit indices suggested that the measurement model demonstrated good fit [c2(df = 693) = 9098.024, p < .001; CFI = .963; TLI = .961; RMSEA = .049]. Total effects analyses revealed that time 1 depression severity significantly predicted greater time 3 GA severity 18 years later (b = 0.158, SE = 0.035 p < .001). Indirect effects analyses showed that time 2 PC (b = 0.047, 99% confidence interval (CI) [0.014, 0.054]), but not PM (b = 0.007, 99% CI [-0.004, 0.010]), mediated the prospective link between depression and GA severity. Higher baseline depression led to lower time 2 PC (b = -0.195, SE = 0.037, p < .001), which in turn uniquely contributed to greater time 3 GA severity (b = -0.175, SE = 0.022, p < .001). Lack of PC functions as both a distal risk factor and salient mechanism by which depression predicts GA close to two decades later.

Hani Zainal

Doctoral Student
The Pennsylvania State University

Presentation(s):

Send Email for Hani Zainal


Assets

Longitudinal Structural Equation Modeling: An Example to Understand the Mechanisms Underlying Depression Predicting Anxiety



Attendees who have favorited this

Please enter your access key

The asset you are trying to access is locked. Please enter your access key to unlock.

Send Email for Longitudinal Structural Equation Modeling: An Example to Understand the Mechanisms Underlying Depression Predicting Anxiety