Grade 25 61 4498 03617 4 g2 1136 33 913 01220 5 g2

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: 0280 26 g2 15.29 10 11.81 1.2140 * 27 g2 15.29 15 19.36 2.7020 * 7 age 56.5 10 15.52 3.1980 * plotfit, uniform=T, branch=.4, compress=T textfit, use.n=T Note that the primary split on grade is the same as when status was used as a dichotomous endpoint, but that the splits thereafter di er. summaryfit,cp=.02 Call: rpartformula = Survpgtime, pgstat ~ age + eet + g2 + grade + gleason + ploidy, data = stagec 1 2 3 4 5 6 7 CP nsplit rel error 0.12913 0 1.0000 0.04169 1 0.8709 0.02880 2 0.8292 0.01720 3 0.8004 0.01518 4 0.7832 0.01271 5 0.7680 0.01000 8 0.7311 xerror 1.0060 0.8839 0.9271 0.9348 0.9647 0.9648 0.9632 xstd 0.07389 0.07584 0.08196 0.08326 0.08259 0.08258 0.08480 Node number 1: 146 observations, complexity param=0.1291 events=54, estimated rate=1 , deviance n=1.338 left son=2 61 obs right son=3 85 obs Primary splits: grade 2.5 to the left, improve=25.270, 0 missing gleason 5.5 to the left, improve=21.630, 3 missing ploidy splits as LRR, improve=14.020, 0 missing g2 13.2 to the left, improve=12.580, 7 missing age 58.5 to the right, improve= 2.796, 0 missing Surrogate splits: gleason 5.5 to the left, agree=0.8630, 0 split ploidy splits as LRR, agree=0.6438, 0 split 42 grade<2.5 | g2<11.36 age>56.5 gleason<5.5 gleason<7.5 3.1980 8/10 0.1220 1/33 g2<13.475 0.5289 4/20 g2>15.29 1.3140 4/8 g2>17.915 0.8817 8/25 1.2140 5/10 0.8795 5/14 2.7020 11/15 2.1830 8/11 Figure 11: The prostate cancer data as a survival tree g2 age 9.945 to the left, agree=0.6301, 0 split 66.5 to the right, agree=0.5890, 0 split . . . Suppose that we wish to simplify this tree, so that only four terminal nodes remain. Looking at the table of complexity parameters, we see that prunefit, cp=.015 would give the desired result. It is also possible to trim the gure interactively using snip.rpart. Point the mouse on a node and click with the left button to get some simple descriptives of the node. Double-clicking with the left button will `remove' the sub-tree below, or one may click on another node. Multiple branches may be snipped o one by one; clicking with the right button will end interactive mode and return the pruned tree. plotfit textfit,use.n=T fit2 - snip.rpartfit node number: 6 n= 75 response= 1.432013  37  43 | grade<2.5 | | g2<11.36 age>56.5 gleason<7.5 3.1980 8/10 gleason<5.5 0.1220 1/33 g2<13.475 0.5289 1.3140 4/20 4/8 0.8817 8/25 g2>15.29 g2>17.915 1.2140 2.7020 5/10 11/15 0.8795 2.1830 5/14 8/11 Figure 12: An illustration of how snip.rpart works. The full tree is plotted in the rst panel. After selecting node 6 with the mouse double clicking on left button, the subtree disappears from the plot shown in the second panel. Finally, the new subtree is redrawn to use all available space and it is labelled. 44 1.0 0.8 0.6 0.4 0.2 Prob Progression 0.0 node node node node 4 5 6 7 0 5 10 15 Time to Progression Figure 13: Survival plot based on snipped rpart object. The probability of t...
View Full Document

Ask a homework question - tutors are online