Working papers:
Working papers:
Miao, W., Beyhum, J., Striaukas, J., & Van Keilegom, I. (2025). High-dimensional censored MIDAS logistic regression for corporate survival forecasting. (R&R) [Replication material]
Xu, H., Miao, W., Dhaene, G., & Beyhum, J. (2026). Bootstrap inference in nonlinear panel data models with interactive fixed effects. [Replication material]
Work in Progress:
Parametric bootstrap inference in panel and network models, with Haoyuan Xu, Geert Dhaene, and Jad Beyhum (Draft available upon request).
This paper proposes a parametric bootstrap approach to bias correction and inference in dense directed network formation models with unobserved heterogeneity. For point estimation, the parametric bootstrap corrects the first-order bias of the maximum likelihood estimator caused by the incidental parameter problem. For inference, it delivers asymptotically valid confidence intervals and hypothesis tests without requiring bias correction. The parametric bootstrap also enables bias-corrected estimation and inference for average partial effects. We illustrate the method using bilateral trade data.
Inference in high-dimensional short panel data after discretizing unobserved heterogeneity, with Jad Beyhum and Artūras Juodis.
Non-linear panel data with time-varying unobserved heterogeneity, with Haoyuan Xu and Jad Beyhum.