start antolini implementation

R/antolini.R

@@ -0,0 +1,45 @@
+cindex = function(pred, meth = c("A", "H"), tiex = 0.5) {
+  times = pred$truth[, 1]
+  outcome = pred$truth[, 2]
+
+  if (meth == "A") {
+    scores = diag(distr6:::C_Vec_WeightedDiscreteCdf(times, sort(unique(times)),
+      cdf = t(1 - pred$data$distr), FALSE, FALSE
+    ))
+  } else {
+    scores = p$data$crank
+  }
+
+  pairs = data.frame(
+    ti = rep(times, length(times)),
+    di = rep(outcome, length(times)),
+    si = rep(scores, length(times)),
+    tj = rep(times, each = length(times)),
+    sj = rep(scores, each = length(times))
+  )
+
+  comparable = function(t_i, t_j, d_i, c) d_i & t_i < t_j & t_i < c
+
+  comp = comparable(pairs$ti, pairs$tj, pairs$di, Inf)
+  if (meth == "A") {
+    conc = pairs[comp, "si"] < pairs[comp, "sj"]
+  } else {
+    conc = pairs[comp, "si"] > pairs[comp, "sj"]
+  }
+  conc = sum(conc) + sum((pairs[comp, "si"] == pairs[comp, "sj"])) * tiex
+
+  conc / sum(comp)
+}
+
+library(mlr3)
+library(mlr3proba)
+set.seed(42)
+t = tsk("rats")
+s = partition(t)
+p = lrn("surv.coxph")$train(t, s$train)$predict(t, s$test)
+p$score(msr("surv.rcll"))
+cindex(p, "A", 0.5)
+
+
+cindex(p, "H", 0.8) - p$score(msr("surv.cindex", tiex = 0.8)) < 1e-6
+microbenchmark::microbenchmark(cindex(p, "H", 0.5), p$score())