luoyuanlab / 3dmice

Missing data imputation for longitudinal multi-variable EHR data. Paper in JAMIA.

3D-MICE: integration of cross-sectional and longitudinal imputation

Requirements

Code is written in R.

Get Started

To train, run (better run as R markdown)

source('tempMICEGPEvalTr.R')

This is a wrapper code calling various subroutines that generate the training data, mask missing values, and performs 3D-MICE imputation, each step is wrapped in its own R source file and should be self-explanatory.

Similarly, to train, run (better run as R markdown)

source('tempMICEGPEvalTe.R')

Citation

@article{luo20173d,
  title={3D-MICE: integration of cross-sectional and longitudinal imputation for multi-analyte longitudinal clinical data},
  author={Luo, Yuan and Szolovits, Peter and Dighe, Anand S and Baron, Jason M},
  journal={Journal of the American Medical Informatics Association},
  volume={25},
  number={6},
  pages={645--653},
  year={2017},
  publisher={Oxford University Press}
}