Factor-augmented transformation models for interval-censored failure time data

Area of Science:

• Biostatistics
• Statistical Modeling
• Data Analysis

Background:

• Interval-censored failure time data are common in research, where exact event times are unknown.
• Multiple correlated covariates can cause multicollinearity, complicating statistical analyses.
• Existing methods may struggle with both interval censoring and high-dimensional correlated predictors.

Purpose of the Study:

• To propose a novel factor-augmented transformation model for interval-censored failure time data.
• To address challenges of dimensionality reduction and multicollinearity in complex datasets.
• To provide a robust statistical framework for analyzing time-to-event data with correlated predictors.

Main Methods:

• Developed a joint modeling framework combining factor analysis and semiparametric transformation models.
• Employed a factor analysis model to group correlated variables into latent factors.
• Utilized a nonparametric maximum likelihood estimation with an expectation-maximization algorithm for implementation.

Main Results:

• The proposed factor-augmented transformation model effectively handles interval-censored data.
• The method successfully reduces dimensionality and mitigates multicollinearity issues.
• Asymptotic properties of estimators were established, and simulation studies confirmed empirical performance.

Conclusions:

• The factor-augmented transformation model offers a powerful approach for analyzing complex failure time data.
• The method is applicable to real-world studies, such as the Alzheimer's Disease Neuroimaging Initiative (ADNI).
• An R package (ICTransCFA) is available for practical application of the proposed methodology.