In the realm of oncology research, emulating clinical trials using
real-world data presents unique challenges, particularly when dealing
with missing data and the need for careful propensity score analyses.
The encore.analytics package addresses these challenges by providing a
comprehensive toolkit that bridges the gap between multiple imputation
and propensity score methodologies.
Real-world oncology data often suffers from missing values, requiring sophisticated imputation techniques. When combined with propensity score analyses for treatment effect estimation, the complexity increases substantially. Traditional approaches often handle these challenges separately, leading to potential biases and methodological inconsistencies.
The encore.analytics package offers an integrated solution that:
- Maintains the proper order of operations between imputation and propensity score estimation
- Ensures consistent handling of uncertainty across multiple imputed datasets
- Provides specialized tools for oncology-specific outcomes and metrics
The package implements various methodologies for:
- Multiple imputation of missing data
- Propensity score estimation and matching
- Survival analysis with multiply imputed data
- Agreement metrics for assessing consistency across imputed datasets
- Specialized visualization tools for survival outcomes
- Cox Model Pooling: Implementation of various approaches for pooling Cox proportional hazards models across multiply imputed datasets
- Kaplan-Meier Pooling: Tools for combining and visualizing Kaplan-Meier curves from multiple imputations
- Raking Weights: Methods for calibrating weights in propensity score analyses
- Agreement Metrics: Functions to assess consistency and agreement across imputed datasets
- Table 1 Generation: Utilities for creating descriptive statistics tables
- Data Simulation: Tools for generating synthetic datasets for testing and validation
You can install the development version of encore.analytics from GitHub with:
# install.packages("pak")
pak::pak("janickweberpals/encore.analytics")For comprehensive documentation and detailed examples, please visit:
- Package Website: https://janickweberpals.github.io/encore.analytics/
- Detailed Workflows: https://janickweberpals.github.io/imputation-ps-workflows/
The documentation includes:
- Background: Theoretical foundation and methodology
- Multiple Survival Model Comparison: Approaches for handling survival analyses with multiple imputations
- Raking Weights: Implementation of weight calibration methods
- Kaplan-Meier Pooling: Techniques for combining survival curves
- Agreement Metrics: Methods for assessing consistency across imputations
The package implements state-of-the-art approaches for:
- Rubin’s rules for combining estimates across imputed datasets
- Various propensity score estimation methods including logistic regression and machine learning approaches
- Specialized pooling methods for survival analyses
- Novel metrics for assessing agreement between imputed datasets
Built with extensibility in mind, encore.analytics features: - A
modular design that allows for easy addition of new methodologies -
Efficient handling of large datasets through optimized algorithms -
Comprehensive unit testing to ensure reliability - Integration with
popular R packages in the survival analysis ecosystem
Contributions are welcome! Please feel free to submit a Pull Request. For major changes, please open an issue first to discuss what you would like to change.
If you use encore.analytics in your research, please cite it as:
Weberpals J (2025). encore.analytics: Multiple Imputation and Propensity Score
Workflows for Oncology Trial Emulation. R package version 0.1.0.
https://github.com/janickweberpals/encore.analytics