README.md

In eric40065/CausalAnalysisforSemiCompRisks: Semi-parametric causal mediation modeling of semicompeting risks

CausalAnalysisforSemiCompRisks

This is an R-package for analyzing semi-competing risks data using causal mediation inference. This package is aimed to illustrate the methodology presented in "Semiparametric causal mediation modeling of hepatitis on mortality through liver cancer incidence".

Installation

devtools::install_github("eric40065/CausalAnalysisforSemiCompRisks")

Usage

The following code can reproduce everything mentioned in our paper. It will take 11 days to reproduce the analyses.

library(CausalAnalysisforSemiCompRisks)
# To reproduce Figure 1(a), 1(b), 1(e), 1(f), 1(i), and 1(j) in page 9
simulation(1, hypo = 'null')
# To reproduce Figure 1(c), 1(d), 1(g), 1(h), 1(k), and 1(l) in page 9
simulation(1, hypo = 'alter')
# To reproduce the first part of Table 1 in page 10
Table11 = simulation(2, hypo = 'null', sample_size = 300)
# To reproduce the second part of Table 1 in page 10
Table12 = simulation(2, hypo = 'null', sample_size = 1000)
# To reproduce the third part of Table 1 in page 10
Table13 = simulation(2, hypo = 'alter', sample_size = 300)
# To reproduce the last part of Table 1 in page 10
Table14 = simulation(2, hypo = 'alter', sample_size = 1000)
# To reproduce Figure 2(a)--(d) in page 11
HBV_result = CASCR(REVEAL_HBV, get_variance = c('asymptotic', 'bootstrap'), plot_result = T)
# To reproduce Figure 1(a)--(j) in page 18 -- 19 of supplement material
HBV_sen_ana = CASCR(REVEAL_HBV, sen_ana = T, get_variance = NULL)

The following code provides a glimpse of the analysis.

simulation

library(CausalAnalysisforSemiCompRisks)
# The result of the unbiasedness as presented in Figure 1 in our paper. We repeat it 1,000
# times to get more accurate result.
simulation(1, hypo = 'null', repeat_size = 100)
simulation(1, hypo = 'alter', repeat_size = 100)

# The result of coverage rate as presented in Table 1 in our paper. We repeat it 1,000 times
# and we set get_variance as c('aymptotic', 'bootstrap') to get more accurate result.
# This, however, spend plenty of time.
simulation(2, hypo = 'null', sample_size = 1000, repeat_size = 100, get_variance = 'asymptotic')
simulation(2, hypo = 'alter', sample_size = 1000, repeat_size = 100, get_variance = 'asymptotic')

REVEAL-HBV

Abstract

• REVEAL-HBV is a community-based prospective cohort study conducted in Taiwan.
• We adjusted age at the cohort entry and the history of alcohol consumption (yes vs. no) as covariates in our model.

Dictionary

REVEAL-HBV contains 11,940 male patients with 8 columns. - hcc.time: time to liver cancer incidence or censored time (days). - die.time: time to death or censored time (days). - hcc.case: whether or not the liver cancer incidence is observed (0: no vs. 1:yes). - dieall.case: whether or not death is observed (no: 0 vs. yes: 1). - HBSAG: hepatitis B surface antigen (negative: 0 vs. positive : 1) - AGE: age at the cohort entry - alcohol1: the alcohol consumption history (no: 0 vs. yes: 1) - smoke1: smoking (no: 0 vs. yes: 1)

Code

library(CausalAnalysisforSemiCompRisks)
# The result of analyzing the REVEAL-HBV dataset as discussed in Section 7 in our paper.
# We set downsample as 1 and get_variance as c('aymptotic', 'bootstrap') in our paper.
result = CASCR(REVEAL_HBV, downsample = 5, get_variance = 'asymptotic', plot_result = T)

# To get a quick result including bootstrap variance, one can run the following code.
# It will take around 15 minutes.
result = CASCR(REVEAL_HBV, downsample = 60, get_variance = c('asymptotic', 'bootstrap'), plot_result = T)

# The result of sensitivity analysis as presented in Section 11 of Supplement Material.
result = CASCR(REVEAL_HBV, sen_ana = T, get_variance = NULL)

More details

More details such as the input data type, parallel programming, and the output can be found using

?CASCR

eric40065/CausalAnalysisforSemiCompRisks documentation built on Dec. 20, 2021, 5:28 a.m.