CompetingRiskSplitFinders: Competing Risk Split Finders
In jatherrien/largeRCRF: Large Random Competing Risks Forests

Description Usage Arguments Details Value Note References Examples

Creates a SplitFinder rJava Java object, which is then used internally when training a competing risk random forest. The split finder is responsible for finding the best split according to the logic of the split finder.

1
2
3

GrayLogRankSplitFinder(events, eventsOfFocus = NULL)

LogRankSplitFinder(events, eventsOfFocus = NULL)

`events`	A vector of integers specifying which competing risk events should be focused on when determining differences. Currently, equal weights will be assigned to all included groups.
`eventsOfFocus`	The split finder will only maximize differences between the two groups with respect to these specified events. Default is `NULL`, which will cause the split finder to focus on all events included in `events`.

These split finders require that the response be CR_Response.

The user only needs to pass this object into train as the splitFinder parameter.

Roughly speaking, the Gray log-rank split finder looks at differences between the cumulative incidence functions of the two groups, while the plain log-rank split finder look at differences between the cause-specific hazard functions. See the references for a more detailed discussion.

An internal rJava Java object used in train.

The Gray log-rank split finder requires that the response include the censoring time.

Kogalur, U., Ishwaran, H. Random Forests for Survival, Regression, and Classification: A Parallel Package for a General Implemention of Breiman's Random Forests: Theory and Specifications. URL https://kogalur.github.io/randomForestSRC/theory.html#section8.2

Ishwaran, H., et. al. (2014) Random survival forests for competing risks, Biostatistics (2014), 15, 4, pp. 757–773