Modeling the effects of a bidirectional latent predictor from multivariate questionnaire data

Area of Science:

• Psychology
• Biostatistics
• Epidemiology

Background:

• Stress is commonly measured using life event questionnaires with ordinal rankings.
• Accurate stress measurement is crucial for understanding its health effects.
• Existing methods may not fully capture the nuances of positive and negative stress reactions.

Purpose of the Study:

• To propose a novel latent variable model for quantifying individual stress.
• To model both negative and positive stress reactions to life events.
• To jointly model stress and health outcomes without assuming event weights.

Main Methods:

• Developed a latent variable model with event-specific negative and positive reaction scores.
• Calculated overall positive and negative stress by summing reactivity scores.
• Employed Bayesian methods and Markov Chain Monte Carlo (MCMC) for joint model fitting.
• Applied the model to investigate the impact of stress on preterm delivery.

Main Results:

• The model successfully differentiates between positive and negative stress responses.
• Event-specific scores inform overall stress levels.
• Joint modeling allowed for inferences without pre-assigned event weights.
• Demonstrated application in predicting preterm delivery based on stress levels.

Conclusions:

• The proposed latent variable model offers a robust framework for stress assessment.
• This approach enhances the understanding of stress-related health outcomes.
• The methodology is applicable to various health research scenarios involving stress.