DM: FYI: Prediction accuracy of S-Plus trees vs. CARTŪ (r)

[Tjen-Sien Lim]

                                Error Rates                    
            ----------------------------------------------------
Datasets    S-Plus tree()   Ripley's Tree()   RPART      CARTŪ (r)
----------------------------------------------------------------

adt             0.145           0.155         0.155      0.142
att             0.380           0.398         0.392      0.394 
ban             0.352           0.372         0.320      0.322
bcw             0.0601          0.0629        0.0630     0.0586
bio             0.150           0.150         0.155      0.164
bld             0.375           0.320         0.328      0.367
bos             0.302           0.275         0.246      0.244
bpr             0.392           0.397         0.418      0.313
cmc             0.483           0.498         0.454      0.456
crx             0.157           0.149         0.159      0.149
der             0.0996          0.0805        0.0622     0.0460
dna             0.0742          0.0742        0.0717     0.0675
ech             0.412           0.347         0.334      0.357
edu             0.446           0.470         0.452      0.436
hab             0.265           0.268         0.284      0.265
hco             0.185           0.152         0.160      0.169
hea             0.208           0.208         0.238      0.221
hep             0.195           0.193         0.225      0.233
hin             0.289           0.290         0.279      0.279
hur             0.213           0.195         0.200      0.152
hyp             0.0111          0.0111        0.00917    0.00727
imp             0.403           0.363         0.369      0.233
inf             0.259           0.235         0.248      0.220
lbw             0.344           0.365         0.359      0.328
led             0.342           0.342         0.343      0.301
pid             0.251           0.245         0.239      0.245
sat             0.234           0.227         0.219      0.152
seg             0.111           0.0983        0.0905     0.0537
smo             0.305           0.305         0.305      0.305
tae             0.655           0.736         0.701      0.351
usn             0.411           0.286         0.274      0.279
veh             0.333           0.325         0.329      0.307
vot             0.0645          0.0622        0.0506     0.0529
wav             0.298           0.319         0.316      0.290  

Mean            0.271           0.264         0.260      0.234


Options used:
1. S-Plus tree():
   - tree.control(minsize=2, mindev=0)
   - predict.tree(..., split=F)

2. Brian Ripley's Tree():
   - tree.control(minsize=2, mindev=0)
   - split="deviance"
   - predict.tree(..., split=T)

3. RPART:
   - rpart.control(minsplit=3, minbucket=1, xval=10)

4. CARTŪ (r):
   - LIMIT ATOM=1

For more details about our project as well as the datasets for
download, go to

   http://www.stat.wisc.edu/~limt/mv.html

-- 
Tjen-Sien Lim                (608) 262-8181                        
Dept. of Statistics          limt@stat.wisc.edu
Univ. of Wisconsin-Madison   http://www.stat.wisc.edu/~limt
1210 West Dayton Street       
Madison, WI 53706