Prior Adaptive Semi-supervised (PASS) Learning of EHR data. PASS is a semi-supervised EHR phenotyping approach using unlabelled data and a surrogate variable to enhance the learning accuracy and efficiency when the size of labels is limited. Compared with existing semi-supervised methods, it is more robust and adaptive to surrogate of poor quality. Reference: https://arxiv.org/abs/2003.11744.
Please see Example.Rmd as a running example for the simulation settings in our paper and the case study with the CAD EHR dataset (CAD_norm_pub.rda) published by https://celehs.github.io/PheCAP/articles/example2.html. All the remaining R scripts are function scripts.