Survival Feature Ranking — survivalfeaturerank • ClinicoPath

Performs univariate survival analysis for multiple features to identify potential prognostic factors. This analysis runs a separate Cox proportional hazards model for each selected feature and ranks them by statistical significance, hazard ratio, or concordance index. Useful for biomarker screening, exploratory analysis, and feature selection before building multivariable models.

Usage

survivalfeaturerank(
  data,
  survtime = NULL,
  event = NULL,
  eventLevel,
  features = NULL,
  rankBy = "pvalue",
  showCI = TRUE,
  adjustPValues = FALSE,
  adjustMethod = "fdr",
  showForestPlot = TRUE,
  forestStyle = "standard",
  showTopKM = FALSE,
  topN = 5,
  kmLayout = "separate",
  endplot = 60,
  byplot = 12,
  risktable = FALSE,
  pplot = TRUE,
  showSummary = TRUE,
  alphaLevel = 0.05,
  showFullTable = TRUE
)

Arguments

data: The dataset to be analyzed, provided as a data frame.
survtime: The numeric variable representing follow-up time until the event or censoring.
event: The event indicator variable (e.g., death, recurrence). Should be coded as 1 = event, 0 = censored, or as a factor where one level represents the event.
eventLevel: The level of the event variable that represents the event of interest. For numeric variables, typically "1". For factors, select the appropriate level.
features: Variables to be tested individually for association with survival. Each variable will be tested in a separate univariate Cox model. Can include both categorical and continuous variables.
rankBy: Criterion for ranking features: - pvalue: Sorts by statistical significance (smallest p-value first) - hazard: Sorts by effect size (hazard ratio furthest from 1) - cindex: Sorts by discriminative ability (highest C-index first)
showCI: If true, displays 95\ in the results table.
adjustPValues: If true, applies multiple testing correction to p-values. Recommended when testing many features to control false discovery rate.
adjustMethod: Method for p-value adjustment. FDR (Benjamini-Hochberg) is recommended for exploratory analysis. Bonferroni is most conservative.
showForestPlot: If true, generates a forest plot showing hazard ratios and confidence intervals for all features, sorted by the ranking criterion.
forestStyle: Visual style for the forest plot.
showTopKM: If true, generates Kaplan-Meier survival curves for the top-ranked features. Useful for visualizing the survival differences for the most important predictors.
topN: Number of top-ranked features for which to generate Kaplan-Meier plots. Only used if "Show Kaplan-Meier Plots for Top Features" is enabled.
kmLayout: Layout for multiple Kaplan-Meier plots. Separate shows one plot per feature, faceted shows all in a grid.
endplot: Maximum follow-up time to display on Kaplan-Meier plots.
byplot: Interval for time axis labels on plots.
risktable: If true, displays the number at risk below Kaplan-Meier plots.
pplot: If true, displays log-rank test p-values on Kaplan-Meier plots.
showSummary: If true, displays summary statistics including total number of features tested, number of significant features, and interpretation guidance.
alphaLevel: Significance level for highlighting significant features in the results table.
showFullTable: If true, shows results for all tested features. If false, shows only statistically significant features.

Value

A results object containing:

`results$instructions`					a html
`results$summaryText`					a preformatted
`results$rankingTable`					a table
`results$forestPlot`					an image
`results$topFeaturesHeading`					a html
`results$kmPlot1`					an image
`results$kmPlot2`					an image
`results$kmPlot3`					an image
`results$kmPlot4`					an image
`results$kmPlot5`					an image
`results$kmPlot6`					an image
`results$kmPlot7`					an image
`results$kmPlot8`					an image
`results$kmPlot9`					an image
`results$kmPlot10`					an image
`results$exportRanking`					an output
`results$interpretation`					a html

Tables can be converted to data frames with asDF or as.data.frame. For example:

results$rankingTable$asDF

as.data.frame(results$rankingTable)

Examples

# Example 1: Basic feature ranking
library(survival)
data(colon)

survivalfeaturerank(
    data = colon,
    survtime = "time",
    event = "status",
    eventLevel = "1",
    features = c("sex", "obstruct", "perfor", "adhere", "nodes", "differ"),
    rankBy = "pvalue"
)

# Example 2: Rank by hazard ratio with plots
survivalfeaturerank(
    data = colon,
    survtime = "time",
    event = "status",
    eventLevel = "1",
    features = c("sex", "obstruct", "perfor", "age", "nodes"),
    rankBy = "hazard",
    showForestPlot = TRUE,
    showTopKM = TRUE,
    topN = 3
)

# Example 3: Rank by C-index for predictive power
survivalfeaturerank(
    data = colon,
    survtime = "time",
    event = "status",
    eventLevel = "1",
    features = c("age", "nodes", "extent", "surg"),
    rankBy = "cindex",
    showCI = TRUE,
    adjustPValues = TRUE,
    adjustMethod = "fdr"
)