This repository contains the code used to replicates examples 1 to 5 contained in the “A computationally efficient nonparametric sampling method of time to event for individual-level models” manuscript (NOTE: ADD DOI AND LINK ONCE PUBLISHED).
This repository also provides a function to draw samples from a
multivariate categorical distribution. This function has an R and a
Python implementations, which are located in
R/nps_nhppp.R
and
python/nps_nhppp.py, respectively.
This repository contains the code used to execute all the examples using
R, which are located inside the
analysis
folder. The examples are the following:
-
01_parametric_hazards.R. Shows how the NPS method can be implemented to draw times to event using discretized hazards derived coming from parametric distributions. -
02_homogeneous_cohort.R. Shows how the NPS method can be implemented to draw times to event using discretized hazards derived life tables from a homogeneous cohort. -
03_heterogeneous_cohort.R. Shows how the NPS method can be implemented to draw times to event using discretized hazards derived life tables from a heterogeneous cohort. -
04_hazards_with_time_dependent_covariates.R. Shows how the NPS method can be implemented to draw times to event using discretized hazards derived from processes following proportional hazards with the next specification$h_i(t) = h_0(t) e^{(x_i(t)\beta)} = h_0(t) e^{((\alpha_0 + \alpha_1 t)\beta)}$ , with$x_i(t) = \alpha_0 + \alpha_1 t$ . -
05_time_dependent_covariates_following_random_paths.R. Shows how the NPS method can be implemented to draw times to event using discretized hazards derived from processes following a parametric baseline hazard with random covariates with the next specification$h_i(t) = h_0(t) e^{(x_i(t)\beta)} = h_0(t) e^{((\alpha_0 + \alpha_1 t)\beta)}$ , with$x_i(t) = \alpha_0 + \alpha_1 y_i(t)$ where$y_i(t)$ follows a Gaussian random walking process$y_i(t) = y_i(t-1) + \epsilon_i$ , and where$\epsilon_i \sim Normal(\mu = 0, \sigma = 0.5)$ .
Finally, we provide a Python implementation for the first three
examples, which are located in the
python
folder.