On board being counted m l in this call the

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: taining the observation time in column 1 and the number of events in column 2, or it may be a vector of event counts alone. fit - rpartskips ~ Opening + Solder + Mask + PadType + Panel, data=solder, method='poisson', control=rpart.controlcp=.05, maxcompete=2 The print command summarizes the tree, as in the previous examples. node, split, n, deviance, yval * denotes terminal node 1 root 900 8788.0 5.530 2 Opening:M,L 600 2957.0 2.553 4 Mask:A1.5,A3,B3 420 874.4 1.033 * 5 Mask:A6,B6 180 953.1 6.099 * 3 Opening:S 300 3162.0 11.480 6 Mask:A1.5,A3 150 652.6 4.535 * 7 Mask:A6,B3,B6 150 1155.0 18.420 * The response value is the expected event rate with a time variable, or in this case the expected number of skips. The values are shrunk towards the global estimate of 5.530 skips observation. The deviance is the same as the null deviance sometimes called the residual deviance that you'd get when calculating a Poisson glm model for the given subset of data. summaryfit,cp=.10 38 Call: rpartformula = skips ~ Opening + Solder + Mask + PadType + Panel, data = solder, method = "poisson", control = rpart.controlcp = 0.05, maxcompete = 2 1 2 3 4 CP nsplit rel error 0.3038 0 1.0000 0.1541 1 0.6962 0.1285 2 0.5421 0.0500 3 0.4137 xerror 1.0051 0.7016 0.5469 0.4187 xstd 0.05248 0.03299 0.02544 0.01962 Node number 1: 900 observations, complexity param=0.3038 events=4977, estimated rate=5.53 , deviance n=9.765 left son=2 600 obs right son=3 300 obs Primary splits: Opening splits as RLL, improve=2670, 0 missing Mask splits as LLRLR, improve=2162, 0 missing Solder splits as RL, improve=1168, 0 missing Node number 2: 600 observations, complexity param=0.1285 events=1531, estimated rate=2.553 , deviance n=4.928 left son=4 420 obs right son=5 180 obs Primary splits: Mask splits as LLRLR, improve=1129.0, 0 missing Opening splits as -RL, improve= 250.8, 0 missing Solder splits as RL, improve= 219.8, 0 missing Node number 3: 300 observations, complexity param=0.1541 events=3446, estimated rate=11.48 , deviance n=10.54 left son=6 150 obs right son=7 150 obs Primary splits: Mask splits as LLRRR, improve=1354.0, 0 missing Solder splits as RL, improve= 976.9, 0 missing PadType splits as RRRRLRLRLL, improve= 313.2, 0 missing Surrogate splits: Solder splits as RL, agree=0.6, 0 split Node number 4: 420 observations events=433, estimated rate=1.033 , deviance n=2.082 Node number 5: 180 observations events=1098, estimated rate=6.099 , deviance n=5.295 Node number 6: 150 observations events=680, estimated rate=4.535 , deviance n=4.351 39 Opening:bc | Mask:abd Opening:bc | Mask:ab Mask:ab 2.553 1531/600 1.033 433/420 6.099 1098/180 4.535 680/150 18.420 2766/150 4.535 680/150 18.420 2766/150 Figure 10: The rst gure shows the solder data, t with the poisson method, using a cp value of 0.05. The second gure shows the same t, but with a cp value of 0.15. The function prune.rpart was used to produce the smaller tree. Node number 7: 150 observations events=2766, estimated rate=18.42 , deviance n=7.701 The improvement is Devian...
View Full Document

Ask a homework question - tutors are online