Competing Risks Cox Learner with Likelihood Based Boosting

Fits a competing risks Cox model using likelihood based boosting. Calls CoxBoost::CoxBoost() or CoxBoost::cv.CoxBoost() from package CoxBoost. Predicts cumulative incidence functions (one CIF matrix per cause, rows = observations, columns = times). If times are not provided for prediction, the unique event times (across causes) from the training set are used by default.

Dictionary

This Learner can be instantiated via lrn():

lrn("cmprsk.coxboost")

Meta Information

• Task type: “cmprsk”

• Predict Types: “cif”

• Feature Types: “integer”, “numeric”

• Required Packages: mlr3, mlr3cmprsk, mlr3extralearners, CoxBoost

Parameters

Id	Type	Default	Levels	Range
cv	logical	-	TRUE, FALSE	-
maxstepno	integer	100		\([0, \infty)\)
K	integer	10		\([2, \infty)\)
type	character	verweij	verweij, naive	-
folds	untyped	NULL		-
minstepno	integer	50		\([0, \infty)\)
start.penalty	numeric	-		\((-\infty, \infty)\)
iter.max	integer	10		\([1, \infty)\)
upper.margin	numeric	0.05		\([0, 1]\)
unpen.index	untyped	-		-
standardize	logical	TRUE	TRUE, FALSE	-
stepno	integer	100		\([0, \infty)\)
penalty	numeric	-		\((-\infty, \infty)\)
criterion	character	pscore	pscore, score, hpscore, hscore	-
stepsize.factor	numeric	1		\((-\infty, \infty)\)
sf.scheme	character	sigmoid	sigmoid, linear	-
pendistmat	untyped	-		-
connected.index	untyped	-		-
x.is.01	logical	FALSE	TRUE, FALSE	-
return.score	logical	TRUE	TRUE, FALSE	-
trace	logical	FALSE	TRUE, FALSE	-
cmprsk	character	-	csh, ccsh	-
coupled.strata	logical	TRUE	TRUE, FALSE	-
at.step	untyped	-		-
times	untyped	-		-

Custom mlr3 parameters

• cv: If TRUE, uses cross-validation with CoxBoost::cv.CoxBoost() to select the optimal number of boosting steps, then refits the final model using CoxBoost::CoxBoost() with this value. Initialized to FALSE by default.

• penalty: If set to "optimCoxBoostPenalty", uses CoxBoost::optimCoxBoostPenalty() to determine the optimal penalty value and number of steps (via CoxBoost::cv.CoxBoost()) for the final model.

Initial parameter values

• cmprsk: Type of competing risks analysis. Initialized to "csh" (cause-specific hazards) which models each cause separately, treating other causes as censored. "ccsh" (combined cause-specific hazards) is another option which combines information across causes for variable selection using a joint criterion.

References

Binder, Harald, Allignol, Arthur, Schumacher, Martin, Beyersmann, Jan (2009). “Boosting for high-dimensional time-to-event data with competing risks.” Bioinformatics, 25(7), 890–896.

See also

• Dictionary of Learners: mlr3::mlr_learners.

• as.data.table(mlr_learners) for a table of available Learners in the running session (depending on the loaded packages).

• Chapter in the mlr3book: https://mlr3book.mlr-org.com/basics.html#learners

• mlr3learners for a selection of recommended learners.

• mlr3cluster for unsupervised clustering learners.

• mlr3pipelines to combine learners with pre- and postprocessing steps.

• mlr3tuning for tuning of hyperparameters, mlr3tuningspaces for established default tuning spaces.

Author

bblodfon

Super classes

mlr3::Learner -> mlr3cmprsk::LearnerCompRisks -> LearnerCompRisksCoxboost

Methods

Public methods

• LearnerCompRisksCoxboost$new()

• LearnerCompRisksCoxboost$selected_features()

• LearnerCompRisksCoxboost$clone()

Inherited methods

• mlr3::Learner$base_learner()
• mlr3::Learner$configure()
• mlr3::Learner$encapsulate()
• mlr3::Learner$format()
• mlr3::Learner$help()
• mlr3::Learner$predict()
• mlr3::Learner$predict_newdata()
• mlr3::Learner$print()
• mlr3::Learner$reset()
• mlr3::Learner$train()

---

Method new()

Creates a new instance of this R6 class.

Usage

LearnerCompRisksCoxboost$new()

---

Method selected_features()

Returns the set of cause-specific selected features which have non-zero coefficients. Calls the internal coef.CoxBoost() function.

Usage

LearnerCompRisksCoxboost$selected_features(at_step = NULL, cause = 1)

Arguments

at_step

(integer(1))
Which boosting step to get the coefficients for. If no step is given (default), the final boosting step is used.

cause

(integer(1))
Integer value indicating the event of interest.

Returns

(character()) vector of feature names.

---

Method clone()

The objects of this class are cloneable with this method.

Usage

LearnerCompRisksCoxboost$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

# Define the Learner
learner = lrn("cmprsk.coxboost", stepno = 10)
print(learner)
#> 
#> ── <LearnerCompRisksCoxboost> (cmprsk.coxboost): Likelihood-based Boosting for C
#> • Model: -
#> • Parameters: cv=FALSE, stepno=10, cmprsk=csh
#> • Packages: mlr3, mlr3cmprsk, mlr3extralearners, and CoxBoost
#> • Predict Types: [cif]
#> • Feature Types: integer and numeric
#> • Encapsulation: none (fallback: -)
#> • Properties: selected_features and weights
#> • Other settings: use_weights = 'use'

# Define a Task
task = tsk("pbc")
task$col_roles$feature = c("age", "chol", "copper", "platelet", "trig",
 "albumin", "alk.phos", "ast", "bili", "protime")

# Stratification based on event
task$set_col_roles(cols = "status", add_to = "stratum")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
#> cause '1':
#> 10 boosting steps resulting in 3 non-zero coefficients  
#> partial log-likelihood: -53.47794 
#> 
#> cause '2':
#> 10 boosting steps resulting in 6 non-zero coefficients  
#> partial log-likelihood: -298.9053 
print(learner$selected_features(cause = 1)) # for cause = 1
#> [1] "age"  "chol" "trig"
print(learner$selected_features(cause = 2)) # for cause = 2
#> [1] "age"      "copper"   "platelet" "albumin"  "bili"     "protime" 

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
#> cmprsk.auc 
#>  0.8861008