Type:	Package
Title:	Benchmarking and Visualization Toolkit for Penalized Cox Models
Version:	6.1.0
Description:	Creates nomogram visualizations for penalized Cox regression models, with the support of reproducible survival model building, validation, calibration, and comparison for high-dimensional data.
License:	GPL-3 | file LICENSE
LazyData:	TRUE
URL:	https://nanx.me/hdnom/, https://github.com/nanxstats/hdnom
BugReports:	https://github.com/nanxstats/hdnom/issues
Depends:	R (≥ 3.5.0)
Imports:	foreach, ggplot2, glmnet, gridExtra, ncvreg, penalized, survival
Suggests:	doParallel, knitr, ragg, rmarkdown
VignetteBuilder:	knitr
Encoding:	UTF-8
RoxygenNote:	7.3.2
NeedsCompilation:	yes
Packaged:	2025-06-09 01:57:43 UTC; nanx
Author:	Nan Xiao [aut, cre], Qing-Song Xu [aut], Miao-Zhu Li [aut], Frank Harrell [ctb] (rms author), Sergej Potapov [ctb] (survAUC author), Werner Adler [ctb] (survAUC author), Matthias Schmid [ctb] (survAUC author)
Maintainer:	Nan Xiao <me@nanx.me>
Repository:	CRAN
Date/Publication:	2025-06-09 04:50:02 UTC

hdnom: Benchmarking and Visualization Toolkit for Penalized Cox Models

Description

Creates nomogram visualizations for penalized Cox regression models, with the support of reproducible survival model building, validation, calibration, and comparison for high-dimensional data.

Author(s)

Maintainer: Nan Xiao me@nanx.me (ORCID)

Authors:

• Qing-Song Xu qsxu@csu.edu.cn

• Miao-Zhu Li miaozhu.li@duke.edu

Other contributors:

• Frank Harrell f.harrell@vanderbilt.edu (rms author) [contributor]

• Sergej Potapov (survAUC author) [contributor]

• Werner Adler (survAUC author) [contributor]

• Matthias Schmid (survAUC author) [contributor]

See Also

Useful links:

• https://nanx.me/hdnom/

• https://github.com/nanxstats/hdnom

• Report bugs at https://github.com/nanxstats/hdnom/issues

Construct nomogram ojects for high-dimensional Cox models

Description

Construct nomograms ojects for high-dimensional Cox models

Usage

as_nomogram(
  object,
  x,
  time,
  event,
  pred.at = NULL,
  fun.at = NULL,
  funlabel = NULL
)

Arguments

Note

The nomogram visualizes the model under the automatically selected "optimal" hyperparameters (e.g. lambda, alpha, gamma).

Examples

data(smart)
x <- as.matrix(smart[, -c(1, 2)])
time <- smart$TEVENT
event <- smart$EVENT
y <- survival::Surv(time, event)

fit <- fit_lasso(x, y, nfolds = 5, rule = "lambda.1se", seed = 1001)

nom <- as_nomogram(
  fit, x, time, event, pred.at = 365 * 2,
  funlabel = "2-Year Overall Survival Probability"
)

print(nom)
plot(nom)

Calibrate high-dimensional Cox models

Description

Calibrate high-dimensional Cox models

Usage

calibrate(
  x,
  time,
  event,
  model.type = c("lasso", "alasso", "flasso", "enet", "aenet", "mcp", "mnet", "scad",
    "snet"),
  alpha,
  lambda,
  pen.factor = NULL,
  gamma,
  lambda1,
  lambda2,
  method = c("fitting", "bootstrap", "cv", "repeated.cv"),
  boot.times = NULL,
  nfolds = NULL,
  rep.times = NULL,
  pred.at,
  ngroup = 5,
  seed = 1001,
  trace = TRUE
)

Arguments

Examples

data("smart")
x <- as.matrix(smart[, -c(1, 2)])
time <- smart$TEVENT
event <- smart$EVENT
y <- survival::Surv(time, event)

# Fit Cox model with lasso penalty
fit <- fit_lasso(x, y, nfolds = 5, rule = "lambda.1se", seed = 1001)

# Model calibration by fitting the original data directly
cal.fitting <- calibrate(
  x, time, event,
  model.type = "lasso",
  alpha = 1, lambda = fit$lambda,
  method = "fitting",
  pred.at = 365 * 9, ngroup = 5,
  seed = 1010
)

# Model calibration by 5-fold cross-validation
cal.cv <- calibrate(
  x, time, event,
  model.type = "lasso",
  alpha = 1, lambda = fit$lambda,
  method = "cv", nfolds = 5,
  pred.at = 365 * 9, ngroup = 5,
  seed = 1010
)

print(cal.fitting)
summary(cal.fitting)
plot(cal.fitting)

print(cal.cv)
summary(cal.cv)
plot(cal.cv)

# # Test fused lasso, SCAD, and Mnet models
# data(smart)
# x = as.matrix(smart[, -c(1, 2)])[1:500, ]
# time = smart$TEVENT[1:500]
# event = smart$EVENT[1:500]
# y = survival::Surv(time, event)
#
# set.seed(1010)
# cal.fitting = calibrate(
#   x, time, event, model.type = "flasso",
#   lambda1 = 5, lambda2 = 2,
#   method = "fitting",
#   pred.at = 365 * 9, ngroup = 5,
#   seed = 1010)
#
# cal.boot = calibrate(
#   x, time, event, model.type = "scad",
#   gamma = 3.7, alpha = 1, lambda = 0.03,
#   method = "bootstrap", boot.times = 10,
#   pred.at = 365 * 9, ngroup = 5,
#   seed = 1010)
#
# cal.cv = calibrate(
#   x, time, event, model.type = "mnet",
#   gamma = 3, alpha = 0.3, lambda = 0.03,
#   method = "cv", nfolds = 5,
#   pred.at = 365 * 9, ngroup = 5,
#   seed = 1010)
#
# cal.repcv = calibrate(
#   x, time, event, model.type = "flasso",
#   lambda1 = 5, lambda2 = 2,
#   method = "repeated.cv", nfolds = 5, rep.times = 3,
#   pred.at = 365 * 9, ngroup = 5,
#   seed = 1010)
#
# print(cal.fitting)
# summary(cal.fitting)
# plot(cal.fitting)
#
# print(cal.boot)
# summary(cal.boot)
# plot(cal.boot)
#
# print(cal.cv)
# summary(cal.cv)
# plot(cal.cv)
#
# print(cal.repcv)
# summary(cal.repcv)
# plot(cal.repcv)

Externally calibrate high-dimensional Cox models

Description

Externally calibrate high-dimensional Cox models

Usage

calibrate_external(
  object,
  x,
  time,
  event,
  x_new,
  time_new,
  event_new,
  pred.at,
  ngroup = 5
)

Arguments

Examples

library("survival")

# Load imputed SMART data
data(smart)
# Use the first 1000 samples as training data
# (the data used for internal validation)
x <- as.matrix(smart[, -c(1, 2)])[1:1000, ]
time <- smart$TEVENT[1:1000]
event <- smart$EVENT[1:1000]

# Take the next 1000 samples as external calibration data
# In practice, usually use data collected in other studies
x_new <- as.matrix(smart[, -c(1, 2)])[1001:2000, ]
time_new <- smart$TEVENT[1001:2000]
event_new <- smart$EVENT[1001:2000]

# Fit Cox model with lasso penalty
fit <- fit_lasso(
  x, Surv(time, event),
  nfolds = 5, rule = "lambda.min", seed = 1001
)

# External calibration
cal.ext <- calibrate_external(
  fit, x, time, event,
  x_new, time_new, event_new,
  pred.at = 365 * 5, ngroup = 5
)

print(cal.ext)
summary(cal.ext)
plot(cal.ext, xlim = c(0.6, 1), ylim = c(0.6, 1))
# # Test fused lasso, MCP, and Snet models
# data(smart)
# # Use first 500 samples as training data
# # (the data used for internal validation)
# x <- as.matrix(smart[, -c(1, 2)])[1:500, ]
# time <- smart$TEVENT[1:500]
# event <- smart$EVENT[1:500]
#
# # Take 1000 samples as external validation data.
# # In practice, usually use data collected in other studies.
# x_new <- as.matrix(smart[, -c(1, 2)])[1001:2000, ]
# time_new <- smart$TEVENT[1001:2000]
# event_new <- smart$EVENT[1001:2000]
#
# flassofit <- fit_flasso(x, survival::Surv(time, event), nfolds = 5, seed = 11)
# scadfit <- fit_mcp(x, survival::Surv(time, event), nfolds = 5, seed = 11)
# mnetfit <- fit_snet(x, survival::Surv(time, event), nfolds = 5, seed = 11)
#
# cal.ext1 <- calibrate_external(
#   flassofit, x, time, event,
#   x_new, time_new, event_new,
#   pred.at = 365 * 5, ngroup = 5)
#
# cal.ext2 <- calibrate_external(
#   scadfit, x, time, event,
#   x_new, time_new, event_new,
#   pred.at = 365 * 5, ngroup = 5)
#
# cal.ext3 <- calibrate_external(
#   mnetfit, x, time, event,
#   x_new, time_new, event_new,
#   pred.at = 365 * 5, ngroup = 5)
#
# print(cal.ext1)
# summary(cal.ext1)
# plot(cal.ext1)
#
# print(cal.ext2)
# summary(cal.ext2)
# plot(cal.ext2)
#
# print(cal.ext3)
# summary(cal.ext3)
# plot(cal.ext3)

Compute Kaplan-Meier estimated survival probabilities for external calibration

Description

Compute Kaplan-Meier estimated survival probabilities for external calibration

Usage

calibrate_external_surv_prob_true(
  pred_prob,
  grp,
  time_new,
  event_new,
  pred.at,
  ngroup
)

Value

list

Compute Kaplan-Meier estimated survival probabilities for calibration

Description

Compute Kaplan-Meier estimated survival probabilities for calibration

Usage

calibrate_surv_prob_true(pred_prob, grp, time, event, pred.at, ngroup)

Value

list

Compare high-dimensional Cox models by model calibration

Description

Compare high-dimensional Cox models by model calibration

Usage

compare_by_calibrate(
  x,
  time,
  event,
  model.type = c("lasso", "alasso", "flasso", "enet", "aenet", "mcp", "mnet", "scad",
    "snet"),
  method = c("fitting", "bootstrap", "cv", "repeated.cv"),
  boot.times = NULL,
  nfolds = NULL,
  rep.times = NULL,
  pred.at,
  ngroup = 5,
  rule = c("lambda.min", "lambda.1se"),
  seed = 1001,
  trace = TRUE
)

Arguments

Examples

data(smart)
x <- as.matrix(smart[, -c(1, 2)])
time <- smart$TEVENT
event <- smart$EVENT

# Compare lasso and adaptive lasso by 5-fold cross-validation
cmp.cal.cv <- compare_by_calibrate(
  x, time, event,
  model.type = c("lasso", "alasso"),
  method = "fitting",
  pred.at = 365 * 9, ngroup = 5, seed = 1001
)

print(cmp.cal.cv)
summary(cmp.cal.cv)
plot(cmp.cal.cv)

Compare high-dimensional Cox models by model validation

Description

Compare high-dimensional Cox models by model validation

Usage

compare_by_validate(
  x,
  time,
  event,
  model.type = c("lasso", "alasso", "flasso", "enet", "aenet", "mcp", "mnet", "scad",
    "snet"),
  method = c("bootstrap", "cv", "repeated.cv"),
  boot.times = NULL,
  nfolds = NULL,
  rep.times = NULL,
  tauc.type = c("CD", "SZ", "UNO"),
  tauc.time,
  rule = c("lambda.min", "lambda.1se"),
  seed = 1001,
  trace = TRUE
)

Arguments

References

Chambless, L. E. and G. Diao (2006). Estimation of time-dependent area under the ROC curve for long-term risk prediction. Statistics in Medicine 25, 3474–3486.

Song, X. and X.-H. Zhou (2008). A semiparametric approach for the covariate specific ROC curve with survival outcome. Statistica Sinica 18, 947–965.

Uno, H., T. Cai, L. Tian, and L. J. Wei (2007). Evaluating prediction rules for t-year survivors with censored regression models. Journal of the American Statistical Association 102, 527–537.

Examples

data(smart)
x <- as.matrix(smart[, -c(1, 2)])[1:1000, ]
time <- smart$TEVENT[1:1000]
event <- smart$EVENT[1:1000]

# Compare lasso and adaptive lasso by 5-fold cross-validation
cmp.val.cv <- compare_by_validate(
  x, time, event,
  model.type = c("lasso", "alasso"),
  method = "cv", nfolds = 5, tauc.type = "UNO",
  tauc.time = seq(0.25, 2, 0.25) * 365, seed = 1001
)

print(cmp.val.cv)
summary(cmp.val.cv)
plot(cmp.val.cv)
plot(cmp.val.cv, interval = TRUE)

Model selection for high-dimensional Cox models with adaptive elastic-net penalty

Description

Automatic model selection for high-dimensional Cox models with adaptive elastic-net penalty, evaluated by penalized partial-likelihood.

Usage

fit_aenet(
  x,
  y,
  nfolds = 5L,
  alphas = seq(0.05, 0.95, 0.05),
  rule = c("lambda.min", "lambda.1se"),
  seed = c(1001, 1002),
  parallel = FALSE
)

Arguments

Examples

data("smart")
x <- as.matrix(smart[, -c(1, 2)])
time <- smart$TEVENT
event <- smart$EVENT
y <- survival::Surv(time, event)

# To enable parallel parameter tuning, first run:
# library("doParallel")
# registerDoParallel(detectCores())
# then set fit_aenet(..., parallel = TRUE).

fit <- fit_aenet(
  x, y,
  nfolds = 3, alphas = c(0.3, 0.7),
  rule = "lambda.min", seed = c(3, 7)
)

nom <- as_nomogram(
  fit, x, time, event,
  pred.at = 365 * 2,
  funlabel = "2-Year Overall Survival Probability"
)

plot(nom)

Model selection for high-dimensional Cox models with adaptive lasso penalty

Description

Automatic model selection for high-dimensional Cox models with adaptive lasso penalty, evaluated by penalized partial-likelihood.

Usage

fit_alasso(
  x,
  y,
  nfolds = 5L,
  rule = c("lambda.min", "lambda.1se"),
  seed = c(1001, 1002)
)

Arguments

Examples

data("smart")
x <- as.matrix(smart[, -c(1, 2)])
time <- smart$TEVENT
event <- smart$EVENT
y <- survival::Surv(time, event)

fit <- fit_alasso(x, y, nfolds = 3, rule = "lambda.min", seed = c(7, 11))

nom <- as_nomogram(
  fit, x, time, event,
  pred.at = 365 * 2,
  funlabel = "2-Year Overall Survival Probability"
)

plot(nom)

Model selection for high-dimensional Cox models with elastic-net penalty

Description

Automatic model selection for high-dimensional Cox models with elastic-net penalty, evaluated by penalized partial-likelihood.

Usage

fit_enet(
  x,
  y,
  nfolds = 5L,
  alphas = seq(0.05, 0.95, 0.05),
  rule = c("lambda.min", "lambda.1se"),
  seed = 1001,
  parallel = FALSE
)

Arguments

Examples

data("smart")
x <- as.matrix(smart[, -c(1, 2)])
time <- smart$TEVENT
event <- smart$EVENT
y <- survival::Surv(time, event)

# To enable parallel parameter tuning, first run:
# library("doParallel")
# registerDoParallel(detectCores())
# then set fit_enet(..., parallel = TRUE).

fit <- fit_enet(
  x, y,
  nfolds = 3, alphas = c(0.3, 0.7),
  rule = "lambda.min", seed = 3
)

nom <- as_nomogram(
  fit, x, time, event,
  pred.at = 365 * 2,
  funlabel = "2-Year Overall Survival Probability"
)

plot(nom)

Model selection for high-dimensional Cox models with fused lasso penalty

Description

Automatic model selection for high-dimensional Cox models with fused lasso penalty, evaluated by cross-validated likelihood.

Usage

fit_flasso(
  x,
  y,
  nfolds = 5L,
  lambda1 = c(0.001, 0.05, 0.5, 1, 5),
  lambda2 = c(0.001, 0.01, 0.5),
  maxiter = 25,
  epsilon = 0.001,
  seed = 1001,
  trace = FALSE,
  parallel = FALSE,
  ...
)

Arguments

Note

The cross-validation procedure used in this function is the approximated cross-validation provided by the penalized package. Be careful dealing with the results since they might be more optimistic than a traditional CV procedure. This cross-validation method is more suitable for datasets with larger number of observations, and a higher number of cross-validation folds.

Examples

data("smart")
x <- as.matrix(smart[, -c(1, 2)])[1:120, ]
time <- smart$TEVENT[1:120]
event <- smart$EVENT[1:120]
y <- survival::Surv(time, event)

fit <- fit_flasso(
  x, y,
  lambda1 = c(1, 10), lambda2 = c(0.01),
  nfolds = 3, seed = 11
)

nom <- as_nomogram(
  fit, x, time, event,
  pred.at = 365 * 2,
  funlabel = "2-Year Overall Survival Probability"
)

plot(nom)

Model selection for high-dimensional Cox models with lasso penalty

Description

Automatic model selection for high-dimensional Cox models with lasso penalty, evaluated by penalized partial-likelihood.

Usage

fit_lasso(x, y, nfolds = 5L, rule = c("lambda.min", "lambda.1se"), seed = 1001)

Arguments

Examples

data("smart")
x <- as.matrix(smart[, -c(1, 2)])
time <- smart$TEVENT
event <- smart$EVENT
y <- survival::Surv(time, event)

fit <- fit_lasso(x, y, nfolds = 3, rule = "lambda.min", seed = 11)

nom <- as_nomogram(
  fit, x, time, event,
  pred.at = 365 * 2,
  funlabel = "2-Year Overall Survival Probability"
)

plot(nom)

Model selection for high-dimensional Cox models with MCP penalty

Description

Automatic model selection for high-dimensional Cox models with MCP penalty, evaluated by penalized partial-likelihood.

Usage

fit_mcp(
  x,
  y,
  nfolds = 5L,
  gammas = c(1.01, 1.7, 3, 100),
  eps = 1e-04,
  max.iter = 10000L,
  seed = 1001,
  trace = FALSE,
  parallel = FALSE
)

Arguments

Examples

data("smart")
x <- as.matrix(smart[, -c(1, 2)])
time <- smart$TEVENT
event <- smart$EVENT
y <- survival::Surv(time, event)

fit <- fit_mcp(x, y, nfolds = 3, gammas = c(2.1, 3), seed = 1001)

nom <- as_nomogram(
  fit, x, time, event,
  pred.at = 365 * 2,
  funlabel = "2-Year Overall Survival Probability"
)

plot(nom)

Model selection for high-dimensional Cox models with Mnet penalty

Description

Automatic model selection for high-dimensional Cox models with Mnet penalty, evaluated by penalized partial-likelihood.

Usage

fit_mnet(
  x,
  y,
  nfolds = 5L,
  gammas = c(1.01, 1.7, 3, 100),
  alphas = seq(0.05, 0.95, 0.05),
  eps = 1e-04,
  max.iter = 10000L,
  seed = 1001,
  trace = FALSE,
  parallel = FALSE
)

Arguments

Examples

data("smart")
x <- as.matrix(smart[, -c(1, 2)])
time <- smart$TEVENT
event <- smart$EVENT
y <- survival::Surv(time, event)

fit <- fit_mnet(
  x, y,
  nfolds = 3,
  gammas = 3, alphas = c(0.3, 0.6, 0.9),
  max.iter = 15000, seed = 1010
)

nom <- as_nomogram(
  fit, x, time, event,
  pred.at = 365 * 2,
  funlabel = "2-Year Overall Survival Probability"
)

plot(nom)

Model selection for high-dimensional Cox models with SCAD penalty

Description

Automatic model selection for high-dimensional Cox models with SCAD penalty, evaluated by penalized partial-likelihood.

Usage

fit_scad(
  x,
  y,
  nfolds = 5L,
  gammas = c(2.01, 2.3, 3.7, 200),
  eps = 1e-04,
  max.iter = 10000L,
  seed = 1001,
  trace = FALSE,
  parallel = FALSE
)

Arguments

Examples

data("smart")
x <- as.matrix(smart[, -c(1, 2)])
time <- smart$TEVENT
event <- smart$EVENT
y <- survival::Surv(time, event)

fit <- fit_scad(
  x, y,
  nfolds = 3, gammas = c(3.7, 5),
  max.iter = 15000, seed = 1010
)

nom <- as_nomogram(
  fit, x, time, event,
  pred.at = 365 * 2,
  funlabel = "2-Year Overall Survival Probability"
)

plot(nom)

Model selection for high-dimensional Cox models with Snet penalty

Description

Automatic model selection for high-dimensional Cox models with Snet penalty, evaluated by penalized partial-likelihood.

Usage

fit_snet(
  x,
  y,
  nfolds = 5L,
  gammas = c(2.01, 2.3, 3.7, 200),
  alphas = seq(0.05, 0.95, 0.05),
  eps = 1e-04,
  max.iter = 10000L,
  seed = 1001,
  trace = FALSE,
  parallel = FALSE
)

Arguments

Examples

data("smart")
x <- as.matrix(smart[, -c(1, 2)])
time <- smart$TEVENT
event <- smart$EVENT
y <- survival::Surv(time, event)

fit <- fit_snet(
  x, y,
  nfolds = 3,
  gammas = 3.7, alphas = c(0.3, 0.8),
  max.iter = 15000, seed = 1010
)

nom <- as_nomogram(
  fit, x, time, event,
  pred.at = 365 * 2,
  funlabel = "2-Year Overall Survival Probability"
)

plot(nom)

Breslow baseline hazard estimator for glmnet objects

Description

Derived from peperr:::basesurv and gbm::basehaz.gbm.

Usage

glmnet_basesurv(time, event, lp, times.eval = NULL, centered = FALSE)

Arguments

Value

list containing cumulative base hazard

Examples

NULL

Compute glmnet predicted survival probabilities for external calibration

Description

Compute glmnet predicted survival probabilities for external calibration

Usage

glmnet_calibrate_external_surv_prob_pred(object, x_tr, x_te, y_tr, pred.at)

Value

list containing predicted survival probability

Compute glmnet predicted survival probabilities for calibration

Description

Compute glmnet predicted survival probabilities for calibration

Usage

glmnet_calibrate_surv_prob_pred(
  x_tr,
  x_te,
  y_tr,
  alpha,
  lambda,
  pen.factor,
  pred.at
)

Value

list containing predicted survival probability

Survival curve prediction for glmnet objects

Description

Derived from c060::predictProb.coxnet

Usage

glmnet_survcurve(object, time, event, x, survtime)

Arguments

Value

list containing predicted survival probabilities and linear predictors for all samples

Examples

NULL

Automatic alpha tuning function by k-fold cross-validation

Description

Automatic alpha tuning function by k-fold cross-validation

Usage

glmnet_tune_alpha(..., alphas, seed, parallel)

Value

best model object and best alpha

Compute external validation measures for glmnet objects

Description

Compute external validation measures for glmnet objects

Usage

glmnet_validate_external_tauc(
  object,
  x_tr,
  x_te,
  y_tr,
  y_te,
  tauc.type,
  tauc.time
)

Value

time-dependent AUC (tAUC) value

Compute validation measures for glmnet objects

Description

Compute validation measures for glmnet objects

Usage

glmnet_validate_tauc(
  x_tr,
  x_te,
  y_tr,
  y_te,
  alpha,
  lambda,
  pen.factor,
  tauc.type,
  tauc.time
)

Value

time-dependent AUC (tAUC) value

Extract information of selected variables from high-dimensional Cox models

Description

Extract the names and type of selected variables from fitted high-dimensional Cox models.

Usage

infer_variable_type(object, x)

Arguments

Value

A list containing the index, name, type and range of the selected variables.

Examples

data("smart")
x <- as.matrix(smart[, -c(1, 2)])
time <- smart$TEVENT
event <- smart$EVENT
y <- survival::Surv(time, event)

fit <- fit_lasso(x, y, nfolds = 3, rule = "lambda.min", seed = 11)
infer_variable_type(fit, x)

Kaplan-Meier plot with number at risk table for internal calibration and external calibration results

Description

Kaplan-Meier plot with number at risk table for internal calibration and external calibration results

Usage

kmplot(
  object,
  group.name = NULL,
  time.at = NULL,
  col.pal = c("JCO", "Lancet", "NPG", "AAAS")
)

Arguments

Examples

data("smart")
# Use the first 1000 samples as training data
# (the data used for internal validation)
x <- as.matrix(smart[, -c(1, 2)])[1:1000, ]
time <- smart$TEVENT[1:1000]
event <- smart$EVENT[1:1000]

# Take the next 1000 samples as external calibration data
# In practice, usually use data collected in other studies
x_new <- as.matrix(smart[, -c(1, 2)])[1001:2000, ]
time_new <- smart$TEVENT[1001:2000]
event_new <- smart$EVENT[1001:2000]

# Fit Cox model with lasso penalty
fit <- fit_lasso(x, survival::Surv(time, event), nfolds = 5, rule = "lambda.min", seed = 11)

# Internal calibration
cal.int <- calibrate(
  x, time, event,
  model.type = "lasso",
  alpha = 1, lambda = fit$lambda,
  method = "cv", nfolds = 5,
  pred.at = 365 * 9, ngroup = 3
)

kmplot(
  cal.int,
  group.name = c("High risk", "Medium risk", "Low risk"),
  time.at = 1:6 * 365
)

# External calibration
cal.ext <- calibrate_external(
  fit, x, time, event,
  x_new, time_new, event_new,
  pred.at = 365 * 5, ngroup = 3
)

kmplot(
  cal.ext,
  group.name = c("High risk", "Medium risk", "Low risk"),
  time.at = 1:6 * 365
)

Kaplan-Meier Plot with Number at Risk Table

Description

Kaplan-Meier Plot with Number at Risk Table

Usage

kmplot_raw(
  fit,
  group.name = NULL,
  time.at = NULL,
  surv.df = NULL,
  col.pal = NULL
)

Arguments

Log-rank test for internal calibration and external calibration results

Description

Log-rank test for internal calibration and external calibration results

Usage

logrank_test(object)

Arguments

Examples

data("smart")
# Use the first 1000 samples as training data
# (the data used for internal validation)
x <- as.matrix(smart[, -c(1, 2)])[1:1000, ]
time <- smart$TEVENT[1:1000]
event <- smart$EVENT[1:1000]

# Take the next 1000 samples as external calibration data
# In practice, usually use data collected in other studies
x_new <- as.matrix(smart[, -c(1, 2)])[1001:2000, ]
time_new <- smart$TEVENT[1001:2000]
event_new <- smart$EVENT[1001:2000]

# Fit Cox model with lasso penalty
fit <- fit_lasso(
  x, survival::Surv(time, event),
  nfolds = 5, rule = "lambda.min", seed = 11
)

# Internal calibration
cal.int <- calibrate(
  x, time, event,
  model.type = "lasso",
  alpha = 1, lambda = fit$lambda,
  method = "cv", nfolds = 5,
  pred.at = 365 * 9, ngroup = 3
)

logrank_test(cal.int)

# External calibration
cal.ext <- calibrate_external(
  fit, x, time, event,
  x_new, time_new, event_new,
  pred.at = 365 * 5, ngroup = 3
)

logrank_test(cal.ext)

Breslow baseline hazard estimator for ncvreg objects

Description

Derived from peperr:::basesurv and gbm::basehaz.gbm.

Usage

ncvreg_basesurv(time, event, lp, times.eval = NULL, centered = FALSE)

Arguments

Value

list containing cumulative base hazard

Examples

NULL

Compute ncvreg predicted survival probabilities for external calibration

Description

Compute ncvreg predicted survival probabilities for external calibration

Usage

ncvreg_calibrate_external_surv_prob_pred(object, x_tr, x_te, y_tr, pred.at)

Value

list containing predicted survival probability

Compute ncvreg predicted survival probabilities for calibration

Description

Compute ncvreg predicted survival probabilities for calibration

Usage

ncvreg_calibrate_surv_prob_pred(
  x_tr,
  x_te,
  y_tr,
  model.type,
  alpha,
  lambda,
  gamma,
  pred.at
)

Value

list containing predicted survival probability

Survival curve prediction for ncvreg objects

Description

Derived from c060::predictProb.coxnet

Usage

ncvreg_survcurve(object, time, event, x, survtime)

Arguments

Value

list containing predicted survival probabilities and linear predictors for all samples

Examples

NULL

Automatic MCP/SCAD gamma tuning function by k-fold cross-validation

Description

Automatic MCP/SCAD gamma tuning function by k-fold cross-validation

Usage

ncvreg_tune_gamma(..., gammas, eps, max.iter, seed, parallel)

Value

best model object and best gamma

Automatic Mnet/Snet gamma and alpha tuning function by k-fold cross-validation

Description

Automatic Mnet/Snet gamma and alpha tuning function by k-fold cross-validation

Usage

ncvreg_tune_gamma_alpha(..., gammas, alphas, eps, max.iter, seed, parallel)

Value

best model object, best gamma, and best alpha

Compute external validation measures for ncvreg model objects

Description

Compute external validation measures for ncvreg model objects

Usage

ncvreg_validate_external_tauc(
  object,
  x_tr,
  x_te,
  y_tr,
  y_te,
  tauc.type,
  tauc.time
)

Value

time-dependent AUC (tAUC) value

Compute validation measures for ncvreg model objects

Description

Compute validation measures for ncvreg model objects

Usage

ncvreg_validate_tauc(
  x_tr,
  x_te,
  y_tr,
  y_te,
  model.type,
  gamma,
  alpha,
  lambda,
  tauc.type,
  tauc.time
)

Value

time-dependent AUC (tAUC) value

Color Palette for AAAS Journals

Description

A 10-color discrete color palette based on the colors used in figures in AAAS journals such as Science.

Usage

palette_aaas()

Color Palette for Journal of Clinical Oncology (JCO)

Description

A 10-color discrete color palette based on the colors used in figures in Journal of Clinical Oncology.

Usage

palette_jco()

Color Palette for Lancet Journals

Description

A 9-color discrete color palette based on the colors used in figures in Lancet Oncology.

Usage

palette_lancet()

Color Palette for NPG Journals

Description

A 10-color discrete color palette based on the colors used in figures in NPG journals such as Nature Reviews Cancer.

Usage

palette_npg()

Breslow baseline hazard estimator for penfit objects

Description

Derived from peperr:::basesurv and gbm::basehaz.gbm.

Usage

penalized_basesurv(time, event, lp, times.eval = NULL, centered = FALSE)

Arguments

Value

list containing cumulative base hazard

Examples

NULL

Compute penfit predicted survival probabilities for external calibration

Description

Compute penfit predicted survival probabilities for external calibration

Usage

penalized_calibrate_external_surv_prob_pred(object, x_tr, x_te, y_tr, pred.at)

Value

list containing predicted survival probability

Compute penfit predicted survival probabilities for calibration

Description

Compute penfit predicted survival probabilities for calibration

Usage

penalized_calibrate_surv_prob_pred(x_tr, x_te, y_tr, lambda1, lambda2, pred.at)

Value

list containing predicted survival probability

Survival curve prediction for penfit objects

Description

Derived from c060::predictProb.coxnet

Usage

penalized_survcurve(object, time, event, x, survtime)

Arguments

Value

list containing predicted survival probabilities and linear predictors for all samples

Examples

NULL

Automatic lambda tuning function for fused lasso by k-fold cross-validation

Description

Automatic lambda tuning function for fused lasso by k-fold cross-validation

Usage

penalized_tune_lambda(..., lambda1, lambda2, seed, trace, parallel)

Value

best model object, best lambda1, and best lambda2

Compute external validation measures for penfit model objects

Description

Compute external validation measures for penfit model objects

Usage

penalized_validate_external_tauc(
  object,
  x_tr,
  x_te,
  y_tr,
  y_te,
  tauc.type,
  tauc.time
)

Value

time-dependent AUC (tAUC) value

Compute validation measures for penfit model objects

Description

Compute validation measures for penfit model objects

Usage

penalized_validate_tauc(
  x_tr,
  x_te,
  y_tr,
  y_te,
  lambda1,
  lambda2,
  tauc.type,
  tauc.time
)

Value

time-dependent AUC (tAUC) value

Plot calibration results

Description

Plot calibration results

Usage

## S3 method for class 'hdnom.calibrate'
plot(
  x,
  xlim = c(0, 1),
  ylim = c(0, 1),
  col.pal = c("JCO", "Lancet", "NPG", "AAAS"),
  ...
)

Arguments

Examples

NULL

Plot external calibration results

Description

Plot external calibration results

Usage

## S3 method for class 'hdnom.calibrate.external'
plot(
  x,
  xlim = c(0, 1),
  ylim = c(0, 1),
  col.pal = c("JCO", "Lancet", "NPG", "AAAS"),
  ...
)

Arguments

Examples

NULL

Plot model comparison by calibration results

Description

Plot model comparison by calibration results

Usage

## S3 method for class 'hdnom.compare.calibrate'
plot(
  x,
  xlim = c(0, 1),
  ylim = c(0, 1),
  col.pal = c("JCO", "Lancet", "NPG", "AAAS"),
  ...
)

Arguments

Examples

NULL

Plot model comparison by validation results

Description

Plot model comparison by validation results

Usage

## S3 method for class 'hdnom.compare.validate'
plot(
  x,
  interval = FALSE,
  col.pal = c("JCO", "Lancet", "NPG", "AAAS"),
  ylim = NULL,
  ...
)

Arguments

Examples

NULL

Plot nomogram objects

Description

Plot nomogram objects

Usage

## S3 method for class 'hdnom.nomogram'
plot(x, ...)

Arguments

Examples

NULL

Plot optimism-corrected time-dependent discrimination curves for validation

Description

Plot optimism-corrected time-dependent discrimination curves for validation

Usage

## S3 method for class 'hdnom.validate'
plot(x, col.pal = c("JCO", "Lancet", "NPG", "AAAS"), ylim = NULL, ...)

Arguments

Examples

NULL

Plot time-dependent discrimination curves for external validation

Description

Plot time-dependent discrimination curves for external validation

Usage

## S3 method for class 'hdnom.validate.external'
plot(x, col.pal = c("JCO", "Lancet", "NPG", "AAAS"), ylim = NULL, ...)

Arguments

Examples

NULL

Make predictions from high-dimensional Cox models

Description

Predict overall survival probability at certain time points from fitted Cox models.

Usage

## S3 method for class 'hdnom.model'
predict(object, x, y, newx, pred.at, ...)

Arguments

Value

A nrow(newx) x length(pred.at) matrix containing overall survival probablity.

Examples

data("smart")
x <- as.matrix(smart[, -c(1, 2)])
time <- smart$TEVENT
event <- smart$EVENT
y <- survival::Surv(time, event)

fit <- fit_lasso(x, y, nfolds = 5, rule = "lambda.1se", seed = 11)
predict(fit, x, y, newx = x[101:105, ], pred.at = 1:10 * 365)

Print calibration results

Description

Print calibration results

Usage

## S3 method for class 'hdnom.calibrate'
print(x, ...)

Arguments

Examples

NULL

Print external calibration results

Description

Print external calibration results

Usage

## S3 method for class 'hdnom.calibrate.external'
print(x, ...)

Arguments

Examples

NULL

Print model comparison by calibration results

Description

Print model comparison by calibration results

Usage

## S3 method for class 'hdnom.compare.calibrate'
print(x, ...)

Arguments

Examples

NULL

Print model comparison by validation results

Description

Print model comparison by validation results

Usage

## S3 method for class 'hdnom.compare.validate'
print(x, ...)

Arguments

Examples

NULL

Print high-dimensional Cox model objects

Description

Print high-dimensional Cox model objects

Usage

## S3 method for class 'hdnom.model'
print(x, ...)

Arguments

Examples

data("smart")
x <- as.matrix(smart[, -c(1, 2)])
time <- smart$TEVENT
event <- smart$EVENT
y <- survival::Surv(time, event)

fit <- fit_lasso(x, y, nfolds = 5, rule = "lambda.1se", seed = 11)
print(fit)

Print nomograms objects

Description

Print nomograms objects

Usage

## S3 method for class 'hdnom.nomogram'
print(x, ...)

Arguments

Examples

NULL

Print validation results

Description

Print validation results

Usage

## S3 method for class 'hdnom.validate'
print(x, ...)

Arguments

Examples

NULL

Print external validation results

Description

Print external validation results

Usage

## S3 method for class 'hdnom.validate.external'
print(x, ...)

Arguments

Examples

NULL

Imputed SMART study data

Description

Imputed SMART study data (no missing values).

Usage

data(smart)

Format

A numeric matrix with 3873 samples, and 29 variables (27 variables + time variable + event variable):

• Demographics

  • SEX - gender

  • AGE - age in years

• Classical risk factors

  • SMOKING - smoking (never, former, current)

  • PACKYRS - in years

  • ALCOHOL - alcohol use (never, former, current)

  • BMI - Body mass index, in kg/m^2

  • DIABETES

• Blood pressure

  • SYSTH - Systolic, by hand, in mm Hg

  • SYSTBP - Systolic, automatic, in mm Hg

  • DIASTH - Diastolic, by hand, in mm Hg

  • DIASTBP - Diastolic, automatic, in mm Hg

• Lipid levels

  • CHOL - Total cholesterol, in mmol/L

  • HDL - High-density lipoprotein cholesterol, in mmol/L

  • LDL - Low-density lipoprotein cholesterol, in mmol/L

  • TRIG - Triglycerides, in mmol/L

• Previous symptomatic atherosclerosis

  • CEREBRAL - Cerebral

  • CARDIAC - Coronary

  • PERIPH - Peripheral

  • AAA - Abdominal aortic aneurysm

• Markers of atherosclerosis

  • HOMOC - Homocysteine, in \mumol/L

  • GLUT - Glutamine, in \mumol/L

  • CREAT - Creatinine clearance, in mL/min

  • ALBUMIN - Albumin (no, micro, macro)

  • IMT - Intima media thickness, in mm

  • STENOSIS - Carotid artery stenosis > 50%

Note

See data-raw/smart.R for the code to generate this data.

References

Steyerberg, E. W. (2008). Clinical prediction models: a practical approach to development, validation, and updating. Springer Science & Business Media.

Examples

data(smart)
dim(smart)

Original SMART study data

Description

Original SMART study data (with missing values) from Steyerberg et, al. 2008.

Usage

data(smarto)

Format

A numeric matrix with 3873 samples, and 29 variables (27 variables + time variable + event variable):

• Demographics

  • SEX - gender

  • AGE - age in years

• Classical risk factors

  • SMOKING - smoking (never, former, current)

  • PACKYRS - in years

  • ALCOHOL - alcohol use (never, former, current)

  • BMI - Body mass index, in kg/m^2

  • DIABETES

• Blood pressure

  • SYSTH - Systolic, by hand, in mm Hg

  • SYSTBP - Systolic, automatic, in mm Hg

  • DIASTH - Diastolic, by hand, in mm Hg

  • DIASTBP - Diastolic, automatic, in mm Hg

• Lipid levels

  • CHOL - Total cholesterol, in mmol/L

  • HDL - High-density lipoprotein cholesterol, in mmol/L

  • LDL - Low-density lipoprotein cholesterol, in mmol/L

  • TRIG - Triglycerides, in mmol/L

• Previous symptomatic atherosclerosis

  • CEREBRAL - Cerebral

  • CARDIAC - Coronary

  • PERIPH - Peripheral

  • AAA - Abdominal aortic aneurysm

• Markers of atherosclerosis

  • HOMOC - Homocysteine, in \mumol/L

  • GLUT - Glutamine, in \mumol/L

  • CREAT - Creatinine clearance, in mL/min

  • ALBUMIN - Albumin (no, micro, macro)

  • IMT - Intima media thickness, in mm

  • STENOSIS - Carotid artery stenosis > 50%

References

Steyerberg, E. W. (2008). Clinical prediction models: a practical approach to development, validation, and updating. Springer Science & Business Media.

Examples

data(smarto)
dim(smarto)

Summary of calibration results

Description

Summary of calibration results

Usage

## S3 method for class 'hdnom.calibrate'
summary(object, ...)

Arguments

Examples

NULL

Summary of external calibration results

Description

Summary of external calibration results

Usage

## S3 method for class 'hdnom.calibrate.external'
summary(object, ...)

Arguments

Examples

NULL

Summary of model comparison by calibration results

Description

Summary of model comparison by calibration results

Usage

## S3 method for class 'hdnom.compare.calibrate'
summary(object, ...)

Arguments

Examples

NULL

Summary of model comparison by validation results

Description

Summary of model comparison by validation results

Usage

## S3 method for class 'hdnom.compare.validate'
summary(object, silent = FALSE, ...)

Arguments

Examples

NULL

Summary of validation results

Description

Summary of validation results

Usage

## S3 method for class 'hdnom.validate'
summary(object, silent = FALSE, ...)

Arguments

Examples

NULL

Summary of external validation results

Description

Summary of external validation results

Usage

## S3 method for class 'hdnom.validate.external'
summary(object, silent = FALSE, ...)

Arguments

Examples

NULL

Plot theme (ggplot2) for hdnom

Description

Plot theme (ggplot2) for hdnom

Usage

theme_hdnom(base_size = 14)

Arguments

Validate high-dimensional Cox models with time-dependent AUC

Description

Validate high-dimensional Cox models with time-dependent AUC

Usage

validate(
  x,
  time,
  event,
  model.type = c("lasso", "alasso", "flasso", "enet", "aenet", "mcp", "mnet", "scad",
    "snet"),
  alpha,
  lambda,
  pen.factor = NULL,
  gamma,
  lambda1,
  lambda2,
  method = c("bootstrap", "cv", "repeated.cv"),
  boot.times = NULL,
  nfolds = NULL,
  rep.times = NULL,
  tauc.type = c("CD", "SZ", "UNO"),
  tauc.time,
  seed = 1001,
  trace = TRUE
)

Arguments

References

Chambless, L. E. and G. Diao (2006). Estimation of time-dependent area under the ROC curve for long-term risk prediction. Statistics in Medicine 25, 3474–3486.

Song, X. and X.-H. Zhou (2008). A semiparametric approach for the covariate specific ROC curve with survival outcome. Statistica Sinica 18, 947–965.

Uno, H., T. Cai, L. Tian, and L. J. Wei (2007). Evaluating prediction rules for t-year survivors with censored regression models. Journal of the American Statistical Association 102, 527–537.

Examples

data(smart)
x <- as.matrix(smart[, -c(1, 2)])[1:500, ]
time <- smart$TEVENT[1:500]
event <- smart$EVENT[1:500]
y <- survival::Surv(time, event)

fit <- fit_lasso(x, y, nfolds = 5, rule = "lambda.1se", seed = 11)

# Model validation by bootstrap with time-dependent AUC
# Normally boot.times should be set to 200 or more,
# we set it to 3 here only to save example running time.
val.boot <- validate(
  x, time, event,
  model.type = "lasso",
  alpha = 1, lambda = fit$lambda,
  method = "bootstrap", boot.times = 3,
  tauc.type = "UNO", tauc.time = seq(0.25, 2, 0.25) * 365,
  seed = 1010
)

# Model validation by 5-fold cross-validation with time-dependent AUC
val.cv <- validate(
  x, time, event,
  model.type = "lasso",
  alpha = 1, lambda = fit$lambda,
  method = "cv", nfolds = 5,
  tauc.type = "UNO", tauc.time = seq(0.25, 2, 0.25) * 365,
  seed = 1010
)

# Model validation by repeated cross-validation with time-dependent AUC
val.repcv <- validate(
  x, time, event,
  model.type = "lasso",
  alpha = 1, lambda = fit$lambda,
  method = "repeated.cv", nfolds = 5, rep.times = 3,
  tauc.type = "UNO", tauc.time = seq(0.25, 2, 0.25) * 365,
  seed = 1010
)

# bootstrap-based discrimination curves has a very narrow band
print(val.boot)
summary(val.boot)
plot(val.boot)

# k-fold cv provides a more strict evaluation than bootstrap
print(val.cv)
summary(val.cv)
plot(val.cv)

# repeated cv provides similar results as k-fold cv
# but more robust than k-fold cv
print(val.repcv)
summary(val.repcv)
plot(val.repcv)
# # Test fused lasso, SCAD, and Mnet models
#
# data(smart)
# x = as.matrix(smart[, -c(1, 2)])[1:500,]
# time = smart$TEVENT[1:500]
# event = smart$EVENT[1:500]
# y = survival::Surv(time, event)
#
# set.seed(1010)
# val.boot = validate(
#   x, time, event, model.type = "flasso",
#   lambda1 = 5, lambda2 = 2,
#   method = "bootstrap", boot.times = 10,
#   tauc.type = "UNO", tauc.time = seq(0.25, 2, 0.25) * 365,
#   seed = 1010)
#
# val.cv = validate(
#   x, time, event, model.type = "scad",
#   gamma = 3.7, alpha = 1, lambda = 0.05,
#   method = "cv", nfolds = 5,
#   tauc.type = "UNO", tauc.time = seq(0.25, 2, 0.25) * 365,
#   seed = 1010)
#
# val.repcv = validate(
#   x, time, event, model.type = "mnet",
#   gamma = 3, alpha = 0.3, lambda = 0.05,
#   method = "repeated.cv", nfolds = 5, rep.times = 3,
#   tauc.type = "UNO", tauc.time = seq(0.25, 2, 0.25) * 365,
#   seed = 1010)
#
# print(val.boot)
# summary(val.boot)
# plot(val.boot)
#
# print(val.cv)
# summary(val.cv)
# plot(val.cv)
#
# print(val.repcv)
# summary(val.repcv)
# plot(val.repcv)

Externally validate high-dimensional Cox models with time-dependent AUC

Description

Externally validate high-dimensional Cox models with time-dependent AUC

Usage

validate_external(
  object,
  x,
  time,
  event,
  x_new,
  time_new,
  event_new,
  tauc.type = c("CD", "SZ", "UNO"),
  tauc.time
)

Arguments

References

Chambless, L. E. and G. Diao (2006). Estimation of time-dependent area under the ROC curve for long-term risk prediction. Statistics in Medicine 25, 3474–3486.

Song, X. and X.-H. Zhou (2008). A semiparametric approach for the covariate specific ROC curve with survival outcome. Statistica Sinica 18, 947–965.

Uno, H., T. Cai, L. Tian, and L. J. Wei (2007). Evaluating prediction rules for t-year survivors with censored regression models. Journal of the American Statistical Association 102, 527–537.

Examples

data(smart)
# Use the first 1000 samples as training data
# (the data used for internal validation)
x <- as.matrix(smart[, -c(1, 2)])[1:1000, ]
time <- smart$TEVENT[1:1000]
event <- smart$EVENT[1:1000]

# Take the next 1000 samples as external validation data
# In practice, usually use data collected in other studies
x_new <- as.matrix(smart[, -c(1, 2)])[1001:2000, ]
time_new <- smart$TEVENT[1001:2000]
event_new <- smart$EVENT[1001:2000]

# Fit Cox model with lasso penalty
fit <- fit_lasso(
  x, survival::Surv(time, event),
  nfolds = 5, rule = "lambda.min", seed = 11
)

# External validation with time-dependent AUC
val.ext <- validate_external(
  fit, x, time, event,
  x_new, time_new, event_new,
  tauc.type = "UNO",
  tauc.time = seq(0.25, 2, 0.25) * 365
)

print(val.ext)
summary(val.ext)
plot(val.ext)

# # Test fused lasso, MCP, and Snet models
# data(smart)
# # Use first 600 samples as training data
# # (the data used for internal validation)
# x <- as.matrix(smart[, -c(1, 2)])[1:600, ]
# time <- smart$TEVENT[1:600]
# event <- smart$EVENT[1:600]
#
# # Take 500 samples as external validation data.
# # In practice, usually use data collected in other studies.
# x_new <- as.matrix(smart[, -c(1, 2)])[1001:1500, ]
# time_new <- smart$TEVENT[1001:1500]
# event_new <- smart$EVENT[1001:1500]
#
# flassofit <- fit_flasso(x, survival::Surv(time, event), nfolds = 5, seed = 11)
# scadfit <- fit_mcp(x, survival::Surv(time, event), nfolds = 5, seed = 11)
# mnetfit <- fit_snet(x, survival::Surv(time, event), nfolds = 5, seed = 11)
#
# val.ext1 <- validate_external(
#   flassofit, x, time, event,
#   x_new, time_new, event_new,
#   tauc.type = "UNO",
#   tauc.time = seq(0.25, 2, 0.25) * 365)
#
# val.ext2 <- validate_external(
#   scadfit, x, time, event,
#   x_new, time_new, event_new,
#   tauc.type = "CD",
#   tauc.time = seq(0.25, 2, 0.25) * 365)
#
# val.ext3 <- validate_external(
#   mnetfit, x, time, event,
#   x_new, time_new, event_new,
#   tauc.type = "SZ",
#   tauc.time = seq(0.25, 2, 0.25) * 365)
#
# print(val.ext1)
# summary(val.ext1)
# plot(val.ext1)
#
# print(val.ext2)
# summary(val.ext2)
# plot(val.ext2)
#
# print(val.ext3)
# summary(val.ext3)
# plot(val.ext3)