Calls flexsurv::flexsurvspline().

• lp is predicted by using an internally defined predict method, see details

• distr is predicted by using an internally defined predict method, see details

• crank is identical to lp

Parameter k is changed to 1 and scale is changed to odds, as these are more in line with the Royston/Parmar proposed models, and the package defaults are equivalent to fitting a parametric model and therefore surv.parametric should be used instead.

If fitting a model with k = 0 then consider using surv.parametric as this is likely to have more optimal results, and has more options for tuning.

## Format

R6::R6Class() inheriting from LearnerSurv.

## Details

The distr prediction is estimated using the fitted custom distributions from flexsurv::flexsurvspline() and the estimated coefficients.

As flexible spline models estimate the baseline hazard as the intercept, the linear predictor, lp, can be calculated as in the classical setting. i.e. For fitted coefficients, $$\beta = (\beta_0,...,\beta_P)$$, and covariates $$X^T = (X_0,...,X_P)^T$$, where $$X_0$$ is a column of $$1$$s: $$lp = \beta X$$.

## Construction

LearnerSurvFlexible$new() mlr_learners$get("surv.flexible")
lrn("surv.flexible")


## Meta Information

• Type: "surv"

• Predict Types: distr, lp, crank

• Feature Types: logical, integer, factor, numeric

• Packages: flexsurv survival distr6

## References

Royston, P. and Parmar, M. (2002). Flexible parametric proportional-hazards and proportional-odds models for censored survival data, with application to prognostic modelling and estimation of treatment effects. Statistics in Medicine, 21(15), 2175-2197. doi: 10.1002/sim.1203 .

Other survival learners: LearnerSurvBlackboost, LearnerSurvCVGlmnet, LearnerSurvCoxPH, LearnerSurvGBM, LearnerSurvGamboost, LearnerSurvGlmboost, LearnerSurvGlmnet, LearnerSurvKaplan, LearnerSurvMboost, LearnerSurvNelson, LearnerSurvParametric, LearnerSurvPenalized, LearnerSurvRandomForestSRC, LearnerSurvRanger, LearnerSurvRpart, LearnerSurvSVM