Skip to contents

High-dimensional mediation analysis for longitudinal mediator and survival outcome data

Source: R/hima_survival_long.R
hima_survival_long.Rd

hima_survival_long estimates and tests high-dimensional longitudinal mediation effects for survival data in a counting process framework.

Usage

hima_survival_long(
  X,
  M,
  tstart,
  tstop,
  status,
  id,
  COV = NULL,
  topN = NULL,
  scale = TRUE,
  Bonfcut = 0.05,
  verbose = FALSE,
  parallel = FALSE,
  ncore = 1
)

Arguments

X

A numeric vector of exposure values (do not use data.frame or matrix).

M

A data.frame or matrix of high-dimensional mediators (rows = observations/intervals, columns = mediators).

tstart

A numeric vector of starting times for each observation/interval (e.g., entry time in a counting-process setup).

tstop

A numeric vector of stopping times for each observation/interval (e.g., event/censoring time in a counting-process setup).

status

A numeric vector of censoring indicators (1 = event, 0 = censored).

id

A vector of subject identifiers (used for clustering/random effects).

COV

A matrix or data.frame of adjusting covariates. Rows represent samples, columns represent variables. Can be NULL.

topN

Integer specifying the number of top mediators retained after sure independent screening (SIS). If NULL (default), topN = ceiling(n/log(n)), where n is the number of unique subjects. When topN exceeds the total number of mediators, all mediators are kept (i.e., the low-dimensional scenario).

scale

Logical. Should the function scale the exposure, mediators, and covariates? Default = TRUE.

Bonfcut

Bonferroni-corrected p value cutoff applied to select significant mediators. Default = 0.05.

verbose

Logical. Should progress messages be printed? Default = FALSE.

parallel

Logical. Enable parallel computing for SIS? Default = FALSE.

ncore

Integer specifying the number of cores to use when parallel = TRUE.

Value

A data.frame containing mediation testing results of significant mediators (joint p-value < Bonfcut).

Index

Mediator name of the selected significant mediator.

alpha_hat

Coefficient estimates for the exposure (X) –> mediator (M) model (adjusted for covariates).

alpha_se

Standard error for alpha_hat.

beta_hat

Coefficient estimates for the mediator (M) –> outcome (Y) model (adjusted for covariates and exposure).

beta_se

Standard error for beta_hat.

IDE

Indirect (mediation) effect estimate, i.e., alpha_hat * beta_hat.

rimp

Relative importance of the mediator.

pmax

joint raw p-value of selected significant mediator (based on Bonferroni method).

References

Liu L, Zhang H, Zheng Y, Gao T, Zheng C, Zhang K, Hou L, Liu L. High-dimensional mediation analysis for longitudinal mediators and survival outcomes. Briefings in Bioinformatics. 2025. DOI: 10.1093/bib/bbaf206. PMID: 40350699 PMCID: PMC12066418

Examples

if (FALSE) { # \dontrun{
data(SurvivalLongData)
pheno_data <- SurvivalLongData$PhenoData
mediator_data <- SurvivalLongData$Mediator

hima_survival_long.fit <- hima_survival_long(
  X = pheno_data$Treatment,
  M = mediator_data,
  tstart = pheno_data$Tstart,
  tstop = pheno_data$Tstop,
  status = pheno_data$Status,
  id = pheno_data$ID,
  COV = pheno_data[, c("Sex", "Age")],
  verbose = TRUE
)
hima_survival_long.fit
} # }