I have created a tree using the party package in R. The tree is fine, but it is very large (27 terminal nodes). Attempts to print it result in unreadable files, with oval overlapping each other.
How can I create something readable?
Can I print subsections of the tree? E.g. a first page with the top few nodes and then subsequent pages starting at granddaughter nodes?
Any other suggestions for visualizing such a tree?
EDIT Stephen asked for a reproducible example; that's reasonable but hard here; the interrelationships of the variables are key to there being a tree, and those are hard (at least for me!) to simulate. Also, some variables are categorical and some continuous. But here is a version of the output from printing the tree structure in R. (I substituted V1 etc for variable names)
1) v1 == {AS, IT, LS, NS}; criterion = 1, statistic = 106.413
2) v2 <= 0; criterion = 1, statistic = 37.039
3) v1 == {NS}; criterion = 1, statistic = 34.458
4)* weights = 316
3) v1 == {AS, IT, LS}
5) V3 <= 2; criterion = 1, statistic = 28.409
6) V4 <= 0; criterion = 0.997, statistic = 15.053
7) v5 == {A: Maste, B: Bache, C: Assoc}; criterion = 0.964, statistic = 15.43
8) V6 <= 24.1068; criterion = 0.98, statistic = 11.242
9)* weights = 259
8) V6 > 24.1068
10)* weights = 886
7) v5 == {D: Plus2}
11)* weights = 38
6) V4 > 0
12) V7 <= 0; criterion = 1, statistic = 22.142
13)* weights = 440
12) V7 > 0
14) V8 <= 1; criterion = 0.999, statistic = 17.06
15)* weights = 88
14) V8 > 1
16)* weights = 59
5) V3 > 2
17)* weights = 100
2) v2 > 0
18)* weights = 41
1) v1 == {, BM, CJ, HS}
19) V4 <= 0; criterion = 1, statistic = 60.5
20) V7 <= 0; criterion = 1, statistic = 41.949
21) V9 <= 0; criterion = 0.985, statistic = 15.936
22)* weights = 376
21) V9 > 0
23) V8 <= 1; criterion = 1, statistic = 30.046
24) V10Wks <= 7.142857; criterion = 1, statistic = 19.078
25) v11 <= 3.5738; criterion = 0.989, statistic = 12.966
26)* weights = 524
25) v11 > 3.5738
27)* weights = 853
24) V10Wks > 7.142857
28)* weights = 27
23) V8 > 1
29) v12 <= 0; criterion = 1, statistic = 27.748
30)* weights = 38
29) v12 > 0
31)* weights = 88
20) V7 > 0
32) V14 <= 0; criterion = 1, statistic = 25.564
33) V8 <= 1; criterion = 0.98, statistic = 13.9
34)* weights = 115
33) V8 > 1
35)* weights = 48
32) V14 > 0
36) V13 <= 2; criterion = 0.983, statistic = 11.504
37)* weights = 96
36) V13 > 2
38)* weights = 91
19) V4 > 0
39) V8 <= 1; criterion = 1, statistic = 25.961
40) V3 <= 0; criterion = 0.999, statistic = 17.093
41) V14 <= 0; criterion = 0.965, statistic = 10.183
42)* weights = 127
41) V14 > 0
43)* weights = 480
40) V3 > 0
44)* weights = 172
39) V8 > 1
45) v15 <= 0; criterion = 0.995, statistic = 14.604
46) V9 <= 0; criterion = 0.987, statistic = 12.104
47) v1 == {HS}; criterion = 1, statistic = 21.895
48)* weights = 43
47) v1 == {BM, CJ}
49) v16 <= 0; criterion = 0.979, statistic = 15.049
50)* weights = 30
49) v16 > 0
51)* weights = 14
46) V9 > 0
52)* weights = 34
45) v15 > 0
53)* weights = 141
I hope that gives some idea of the structure; lots of nodes!
By default, plot in party puts each split in an ellipse and puts additional info for the terminal nodes. But that won't fit on a page.
