A longitudinal measurement error model with a semicontinuous covariate.

Area of Science:

• Statistics
• Epidemiology
• Sleep Medicine

Background:

• Covariate measurement error in regression typically assumes additive or multiplicative models.
• These standard assumptions are inadequate for sleep-disordered breathing (SDB) severity measurement in the Wisconsin Sleep Cohort Study (WSCS).
• The observed surrogate for SDB severity (breathing pauses) has a unique nonnegative, semicontinuous distribution with a zero point mass.

Purpose of the Study:

• To propose and implement a novel latent variable measurement error model for SDB.
• To address the specific distributional characteristics of SDB surrogate measures.
• To improve the accuracy of regression analyses involving SDB severity.

Main Methods:

• Development of a latent variable measurement error model tailored for SDB.
• Implementation within a linear mixed model framework.
• Estimation procedure adapted from regression calibration, incorporating distributional assumptions for the latent variable.

Main Results:

• The proposed latent variable model effectively handles the non-standard measurement error structure of SDB.
• The methodology provides a more accurate representation of SDB severity compared to traditional approaches.
• The model was successfully illustrated using data from the Wisconsin Sleep Cohort Study.

Conclusions:

• A latent variable measurement error model is suitable for analyzing SDB severity with non-standard error structures.
• This approach enhances the reliability of statistical inferences in sleep research.
• The findings offer a valuable tool for epidemiological studies on sleep-disordered breathing.