In celehs/PETLER: Multi-modal Automated Survival Time Annotation

Overview

The MASTA algorithm is a semi-supervised learning method. The implementation has 2 steps. In Step I, longitudinal and follow_up_time will be used to extract features from estimated subject-specific intensity functions of individual encounters. In Step II, survival and follow_up_time will be used to train and evaluate risk prediction models with survival outcomes.

For the details of the data preparation, please see the vignette for it.

Installation

Load the package into R.

library(MASTA)

Data Example

head(longitudinal)
table(longitudinal$code)

head(follow_up_time) ; 
nrow(follow_up_time) ;

head(survival)
nrow(survival)

Run Step I (feature selections)

system.time(Z <- fpca.combine(longitudinal, follow_up_time, K.select = "PropVar"))

Run Step II (fitting)

Using the object derived from fpca.combine and the two data files -- survival and follow_up_time

system.time(b <- masta.fit(Z, survival, follow_up_time, Tend=1, cov_group = NULL, thresh = 0.7, PCAthresh = 0.9, seed = 100))
names(b)

celehs/PETLER documentation built on Sept. 3, 2021, 8:21 a.m.