Integration method for time-dependent measures

[mikolajsp]

At this time, for calculation of Integrated Brier Score and Integrated AUC is performed as Riemann integral. This leads to a great influence of times at the end of the considered time-frame (where there might be few observations) on the integrated score.

Two proposed normalization methods are in [1]. We should try to implement (one of/both of) these methods.

There should also be a possibility to cut off the integration at different quantiles of times in the data (Q25, Q50, Q75) [2]

• [1] E. Graf, C. Schmoor, W. Sauerbrei, and M. Schumacher, “Assessment and comparison of prognostic classification schemes for survival data,” Statistics in Medicine, vol. 18, no. 17-18, pp. 2529–2545, 1999.
• [2] P. Kopper, S. Wiegrebe, B. Bischl, A. Bender, D. Rügamer, “DeepPAMM: Deep Piecewise Exponential Additive Mixed Models for Complex Hazard Structures in Survival Analysis”. Advances in Knowledge Discovery and Data Mining. PAKDD 2022.

[hbaniecki]

Also, later use it to update the choice of max_vars in feature importance.

survex/R/plot_surv_feature_importance.R

Line 55 in 561b62e

subs <- aggregate(plotting_df$value, by = list(var = plotting_df$ind), function(x) sum(abs(x)))