Person-time generalized linear regression mixed models in survival analysis

This study presents a methodological contribution within a discrete-time survival analysis framework by introducing a person-time generalized linear mixed model. An application of this methodology, using logistic models with random effects, is provided in the context of academic career progression in Italy, a field where gender disparities remain a pressing issue. The model investigates the timing and progression of academic promotions using a personyear approach. Temporal dynamics and the effect of key (possibly time-varying) predictors are captured through a longitudinal structure of the data. A random effect is included to account for unobserved individual heterogeneity. This approach is designed to capture individual variability and unobserved heterogeneity, which are crucial for understanding career dynamics. Compared to traditional Cox models and logistic models without random effects, our approach offers a more accurate representation of the progression over time and highlights persistent gender inequalities. Furthermore, the model's performance is tested using a second dataset from existing research and simulated data. The possibility to identify heterogeneity also opens the door to future investigations into unobserved factors. This methodological framework provides an adaptable tool for studying complex time-to-event data across various fields, improving both precision and interpretability of the results.