Causal InferenceTransportability/Generalizability Heterogeneous treatment effect Time-dependent exposures and outcomes Non/semi-parametric theory Failure time outcomes |
Machine LearningConvergence rate characterization A prior knowledge incorporation Structured learning Mixed integer programming High-dimensional data |
ApplicationCardiovascular disease Infectious disease Mental health Oncology HIV |

1.2:

1.3:

1.4:

1.5:

1.6:

1.7:

1.8:

1.9: A massive simulation comparing 16 parametric and Bayesian methods for the use of hybrid control in early phase umbrella

1.10: Estimate the heterogeneous treatment effects with

1.11: Develop a

1.12:

We strongly encourage researchers to thoroughly understand and analyze the data before applying any analytical methods. Integrating this information can significantly enhance the performance of the model in use.

For example, when it comes to variable selection, incorporating specific selection rules can lead to improvements in prediction accuracy, a reduced false alarm rate, and, notably, enhanced interpretability of the selected model.

Examples of simple rules can be "if the interaction is chosen, then all or at least one of the main terms must also be selected", "if the subtopic is selected, then the overarching topic should also be chosen", "select at least one gene from each pathway", and "collectively select dummy variables representing a categorical variable."

We have developed unified methods for systematically integrating all available information into the analysis process.

2.1 Developed a formal mathematical language for expressing all possible selection dependencies that can be incorporated into structured variable selection, and provide the algorithms for identifying all permissible variable subsets that satisfying a given selection rule.

2.2 Specifically, for the latent overlapping group, we develop roadmaps of grouping structure identification and apply it to identify predictors of major bleeding among hospitalized hypertensive patients using oral anticoagulants for atrial fibrillation, where seven rules are being considered.

2.3 Methods for incorporating flexible selection rules into time-dependent Cox models.

2.5 Application of variable selection in HIV