Re: [R] ctree (party) - select a specific variable in the first split
Hi Achim, Thank you for the quick suggestion. I actually did the first split manually, then grew two trees in the two subsets, it worked. Best, Sal Hello, My question is related to ctree() function from the library 'party'. See fortune(58). Yeah, package 'party' - sorry for the confusion. Is there a way to force ctree() to use a specific variable in the first split? No way that does this automatically. I am asking because the first split contains two variables with very similar scores, and choosing the alternative variable would induce a tree somewhat easier to interpret. If both variables are closely correlated and you believe one of the variables to be more meaningful for partitioning, then omit the other from tree construction. Alternatively, you can grow two trees in the two subsets based on the first variable. hth, Z Thanks, Sal __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. -- Salvatore Loguercio, Ph.D. Research Associate - Su Lab Department of Molecular and Experimental Medicine The Scripps Research Institute http://sulab.org logue...@scripps.edu 858-784-2112 __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] ctree (party) - select a specific variable in the first split
On Mon, 1 Apr 2013, Salvatore Loguercio wrote: Hello, My question is related to ctree() function from the library 'party'. See fortune(58). Is there a way to force ctree() to use a specific variable in the first split? No way that does this automatically. I am asking because the first split contains two variables with very similar scores, and choosing the alternative variable would induce a tree somewhat easier to interpret. If both variables are closely correlated and you believe one of the variables to be more meaningful for partitioning, then omit the other from tree construction. Alternatively, you can grow two trees in the two subsets based on the first variable. hth, Z Thanks, Sal __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] ctree
Did you try the example that I sent you? I think it does what you ask for... (sent from mobile phone)mia88 wrote:Hey thanks for your help. Im afraid its a little bit too complicated for me.. Before I do al lot of research in the next days to finally understand it I would like to make sure that it does help me with my problem. Maybe I explain it once again : For example in the first node I can read that group 2 is selected between 0.3 and 0.4 but I need the exact number so that I can use it to get the sensitivity and specifity... -- View this message in context: http://r.789695.n4.nabble.com/ctree-tp4649478p4649538.html Sent from the R help mailing list archive at Nabble.com. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] ctree
Hey thanks for your help. Im afraid its a little bit too complicated for me.. Before I do al lot of research in the next days to finally understand it I would like to make sure that it does help me with my problem. Maybe I explain it once again : For example in the first node I can read that group 2 is selected between 0.3 and 0.4 but I need the exact number so that I can use it to get the sensitivity and specifity... -- View this message in context: http://r.789695.n4.nabble.com/ctree-tp4649478p4649538.html Sent from the R help mailing list archive at Nabble.com. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] ctree
On Wed, 14 Nov 2012, mia88 wrote:
Hello, I plotted a nice tree with "ctree" . It shows 3 nodes with the
prediction of my 2 groups. (see picture) Unfortunately I need a larger
scale to read the exact prediction of my groups to get the specificity
and sensitivity. I tried to change the scale with "axis" but it didn't
work, my guess because it's not a normal graph with x and y axis. Has
someone an idea how to change the scales in the nodes of my tree?
I'm not exactly sure how you would like to improve the visualization. I
could think of two options: (1) Use a y-axis labeling with only the
predicted probability (rounded to some precision). (2) Add some label,
e.g., at the bottom of the bars, with the predicted probability.
While both options are in principle possible, neither is available out of
the box. It requires some programming using the "grid" package in which
the party plots are created. However, if you start out from the
node_barplot() function provided in the package, not a lot of programming
is necessary. As an example, I have created a function node_barplot2()
with a modified grid.yaxis() call to implement option (1) outlined above.
See the code below. To try it out, please source the entire function
_including_ the class() assignment at the end. And then you can do:
## package and tree for Pima Indians Diabetes data
library("party")
data("PimaIndiansDiabetes", package = "mlbench")
ct <- ctree(diabetes ~ ., data = PimaIndiansDiabetes)
## visualizations: default and alternative y-axis and more spaces
plot(ct)
plot(ct, terminal_panel = node_barplot2)
plot(ct, terminal_panel = node_barplot2,
tp_args = list(ylines = c(2, 4)))
Hope that helps,
Z
node_barplot2 <- function(ctreeobj,
col = "black",
fill = NULL,
beside = NULL,
ymax = NULL,
ylines = NULL,
widths = 1,
gap = NULL,
reverse = NULL,
id = TRUE)
{
getMaxPred <- function(x) {
mp <- max(x$prediction)
mpl <- ifelse(x$terminal, 0, getMaxPred(x$left))
mpr <- ifelse(x$terminal, 0, getMaxPred(x$right))
return(max(c(mp, mpl, mpr)))
}
y <- response(ctreeobj)[[1]]
if(is.factor(y) || class(y) == "was_ordered") {
ylevels <- levels(y)
if(is.null(beside)) beside <- if(length(ylevels) < 3) FALSE else TRUE
if(is.null(ymax)) ymax <- if(beside) 1.1 else 1
if(is.null(gap)) gap <- if(beside) 0.1 else 0
} else {
if(is.null(beside)) beside <- FALSE
if(is.null(ymax)) ymax <- getMaxPred(ctreeobj@tree) * 1.1
ylevels <- seq(along = ctreeobj@tree$prediction)
if(length(ylevels) < 2) ylevels <- ""
if(is.null(gap)) gap <- 1
}
if(is.null(reverse)) reverse <- !beside
if(is.null(fill)) fill <- gray.colors(length(ylevels))
if(is.null(ylines)) ylines <- if(beside) c(3, 4) else c(1.5, 2.5)
### panel function for barplots in nodes
rval <- function(node) {
## parameter setup
pred <- node$prediction
if(reverse) {
pred <- rev(pred)
ylevels <- rev(ylevels)
}
np <- length(pred)
nc <- if(beside) np else 1
fill <- rep(fill, length.out = np)
widths <- rep(widths, length.out = nc)
col <- rep(col, length.out = nc)
ylines <- rep(ylines, length.out = 2)
gap <- gap * sum(widths)
yscale <- c(0, ymax)
xscale <- c(0, sum(widths) + (nc+1)*gap)
top_vp <- viewport(layout = grid.layout(nrow = 2, ncol = 3,
widths = unit(c(ylines[1], 1, ylines[2]), c("lines", "null",
"lines")),
heights = unit(c(1, 1), c("lines", "null"))),
width = unit(1, "npc"),
height = unit(1, "npc") - unit(2, "lines"),
name = paste("node_barplot", node$nodeID, sep = ""))
pushViewport(top_vp)
grid.rect(gp = gpar(fill = "white", col = 0))
## main title
top <- viewport(layout.pos.col=2, layout.pos.row=1)
pushViewport(top)
mainlab <- paste(ifelse(id, paste("Node", node$nodeID, "(n = "), "n =
"),
sum(node$weights), ifelse(id, ")", ""), sep = "")
grid.text(mainlab)
popViewport()
plot <- viewport(layout.pos.col=2, layout.pos.row=2,
xscale=xscale, yscale=yscale,
name = paste("node_barplot", node$nodeID, "plot",
sep = ""))
pushViewport(plot)
if(beside) {
xcenter <- cumsum(widths+gap) - widths/2
for (i in 1:np) {
grid.rect(x = xcenter[i], y = 0, height = pred[i],
width = widths[i],
just = c("center", "bottom"), default.units
Re: [R] Ctree plot
Hi, You may want to check grid.arrange in gridExtra package. Weidong On Thu, Aug 9, 2012 at 7:36 AM, rodrock wrote: > HI everybody! > > Has anybody figure out how would be possible to plot several ctree plots > beside each other? > I would really appreciate it > My idea would be like this, but the par() function seem to be lower in > hierarchy > Thanks in advance > # > library(party) > > mtree <- ctree(Species ~ ., data=iris) > sep.tree<-ctree(Sepal.Length ~ ., data=iris) > > par(mfrow=c(1,2)) # Does not work > plot(mtree) > plot(sep.tree) > > > > > - > Rodrigo Vargas G. > - > Silviculture Institute > Freiburg University > > -- > View this message in context: > http://r.789695.n4.nabble.com/Ctree-plot-tp4639750.html > Sent from the R help mailing list archive at Nabble.com. > > __ > R-help@r-project.org mailing list > https://stat.ethz.ch/mailman/listinfo/r-help > PLEASE do read the posting guide http://www.R-project.org/posting-guide.html > and provide commented, minimal, self-contained, reproducible code. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] ctree for suvival analysis problem
Thanks David!!! Sorry for the stupid question... Seems like I still have issues with pec, but I will work on it some more... -- View this message in context: http://r.789695.n4.nabble.com/ctree-for-suvival-analysis-problem-tp4630362p4630408.html Sent from the R help mailing list archive at Nabble.com. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] ctree for suvival analysis problem
On May 17, 2012, at 11:04 AM, ofraam wrote: Thanks very much Achim! I was indeed using "pec" which requires "rms"... I see now that the old version of pec that I used didn't require rms and therefore I had no problem.. I'm guessing I don't need the rms methods for my use of pec, but am not sure how I can remove the dependency or alternatively mask back the "Surv" object to survival package. Any ideas for that? Perhaps: GBSG2ct <- ctree(survival::Surv(time, cens) ~ .,data = GBSG2) -- David Winsemius, MD West Hartford, CT __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] ctree for suvival analysis problem
Thanks very much Achim! I was indeed using "pec" which requires "rms"... I see now that the old version of pec that I used didn't require rms and therefore I had no problem.. I'm guessing I don't need the rms methods for my use of pec, but am not sure how I can remove the dependency or alternatively mask back the "Surv" object to survival package. Any ideas for that? Thanks again! Ofra -- View this message in context: http://r.789695.n4.nabble.com/ctree-for-suvival-analysis-problem-tp4630362p4630403.html Sent from the R help mailing list archive at Nabble.com. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] ctree for suvival analysis problem
On Thu, 17 May 2012, ofraam wrote:
Hi All,
I'm using the party package to grow conditional inference trees for survival
analysis.
When I used party version party_0.9-9991 everything worked well, but when I
update to party_1.0-2 (due to using 64bit R), I get an error. For simplicity
I will show the error I get for the example in the party documentation:
### survival analysis
if (require("ipred")) {
data("GBSG2", package = "ipred")
GBSG2ct <- ctree(Surv(time, cens) ~ .,data = GBSG2)
plot(GBSG2ct)
treeresponse(GBSG2ct, newdata = GBSG2[1:2,])
}
Works for me without a problem.
for the plot(GBSG2ct) line I get the following error:
Error in Summary.Surv(c(1814, 2018, 712, 1807, 772, 448, 2172, 2161, 471, :
Invalid operation on a survival time
and for treeresponse(GBSG2ct, newdata = GBSG2[1:2,]) I get this error:
Error: extends(class(y), "Surv") is not TRUE
(I care more about treeresponse but assume it is related).
My guess is that you have loaded package "rms" but haven't told us so.
This overwrites Surv() in a way that is incompatible with the usage in
ctree().
Best,
Z
Any help would be highly appreciated!
Thanks,
Ofra
--
View this message in context:
http://r.789695.n4.nabble.com/ctree-for-suvival-analysis-problem-tp4630362.html
Sent from the R help mailing list archive at Nabble.com.
__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
Re: [R] ctree question
On Wed, 18 Jan 2012, seanstcl...@verizon.net wrote:
Hello. I have used the "party" package to generate a regression tree as
follows:
>origdata<-read.csv("origdata.csv")
>ctrl<-ctree_control(mincriterion=0.99,maxdepth=10,minbucket=10)
>test.ct<-ctree(Y~X1+X2+X3,data=origdata,control=ctrl)
The above works fine. Orig data was my training data. I now have a test
data file (testdata), and would like to run the testdata through the above
tree to see predictions. I tried using the following function
>predict(test.ct,newdata=testdata)
but I get the following error:
Error in checkData(oldData, RET) :
Levels in factors of new data do not match original data
I've looked at the testdata file closely and it does not appear to contain
any levels of factors that were not in the original.
Did you check carefully enough? compare
levels(origdata$X1)
levels(testdata$X1)
and so on for X2 and X3. Maybe the _observed_ levels in the testdata are
ok but not all levels.
Z
What might I be doing
incorrectly, and how can i use the tree that was generated above to
generate predictions for this new file testdata?
THanks.
sean
__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
Re: [R] ctree and survival problem
Yup, thats the culprit. Thx.
//M
On 28. apr. 2011, at 18.36, Achim Zeileis wrote:
> On Thu, 28 Apr 2011, moleps wrote:
>
>> sessionInfo yields the following:
>
> OK, the "Design" package causes the problem here. When you load the Design
> package, it provides a new Surv() and related methods. This clashes with the
> computations of ctree() based on Surv(). So it's better not to load both
> packages simultaneously...
> Z
>
>>> sessionInfo()
>> R version 2.11.1 (2010-05-31)
>> x86_64-apple-darwin9.8.0
>>
>> locale:
>> [1] en_US.UTF-8/en_US.UTF-8/C/C/en_US.UTF-8/en_US.UTF-8
>>
>> attached base packages:
>> [1] stats4tcltk splines grid stats graphics grDevices
>> utils datasets methods base
>>
>> other attached packages:
>> [1] ipred_0.8-8class_7.3-3mlbench_2.1-0 rpart_3.1-46
>>party_0.9-1vcd_1.2-9 colorspace_1.0-1
>> strucchange_1.4-3 coin_1.0-18mvtnorm_0.9-95
>> [11] modeltools_0.2-17 Design_2.3-0 car_2.0-8 nnet_7.3-1
>> sandwich_2.2-6 boot_1.2-43Amelia_1.2-18 lmtest_0.9-27
>>zoo_1.6-4 MASS_7.3-7
>> [21] mgcv_1.7-2 cem_1.0.142randomForest_4.6-2
>> lattice_0.19-17nlme_3.1-97rgl_0.92.798 Hmisc_3.8-3
>> survival_2.36-2ggplot2_0.8.9 proto_0.3-8
>> [31] reshape_0.8.3 plyr_1.4 foreign_0.8-41
>>
>> loaded via a namespace (and not attached):
>> [1] cluster_1.13.2 digest_0.4.2 Matrix_0.999375-46 tools_2.11.1
>>
>>
>> As i described in the original post I ran
>>
>>> example(ctree)
>>
>> and the following part yields an error:
>>
>>
>> if (require("ipred")) {
>> data("GBSG2", package = "ipred")
>> GBSG2ct <- ctree(Surv(time, cens) ~ .,data = GBSG2)
>> plot(GBSG2ct)
>> treeresponse(GBSG2ct, newdata = GBSG2[1:2,])
>> }
>>
>> Error in Summary.Surv(c(1814, 2018, 712, 1807, 772, 448, 2172, 2161, 471, :
>> Invalid operation on a survival time
>>
>>
>> //M
>>
>>
>>
>>
>>
>> On 28. apr. 2011, at 14.19, Jonathan Daily wrote:
>>
>>> It would help people who know more about R's guts than me if you
>>> posted your sessionInfo() output and exactly what commands produced
>>> your error.
>>>
>>> It is also recommended that you try simply upgrading R to the latest
>>> version and see if you get an error with the latest version of
>>> 'party'. My guess is that the error will go away.
>>>
>>> On Wed, Apr 27, 2011 at 3:40 PM, moleps wrote:
Forgot to mention that the ctree command is from the party library.
//M
__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
>>>
>>>
>>>
>>> --
>>> ===
>>> Jon Daily
>>> Technician
>>> ===
>>> #!/usr/bin/env outside
>>> # It's great, trust me.
>>
>> __
>> R-help@r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
>> and provide commented, minimal, self-contained, reproducible code.
>>
__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
Re: [R] ctree and survival problem
On Thu, 28 Apr 2011, moleps wrote:
sessionInfo yields the following:
OK, the "Design" package causes the problem here. When you load the Design
package, it provides a new Surv() and related methods. This clashes with
the computations of ctree() based on Surv(). So it's better not to load
both packages simultaneously...
Z
sessionInfo()
R version 2.11.1 (2010-05-31)
x86_64-apple-darwin9.8.0
locale:
[1] en_US.UTF-8/en_US.UTF-8/C/C/en_US.UTF-8/en_US.UTF-8
attached base packages:
[1] stats4tcltk splines grid stats graphics grDevices utils
datasets methods base
other attached packages:
[1] ipred_0.8-8class_7.3-3mlbench_2.1-0 rpart_3.1-46
party_0.9-1vcd_1.2-9 colorspace_1.0-1 strucchange_1.4-3
coin_1.0-18mvtnorm_0.9-95
[11] modeltools_0.2-17 Design_2.3-0 car_2.0-8 nnet_7.3-1
sandwich_2.2-6 boot_1.2-43Amelia_1.2-18 lmtest_0.9-27
zoo_1.6-4 MASS_7.3-7
[21] mgcv_1.7-2 cem_1.0.142randomForest_4.6-2 lattice_0.19-17
nlme_3.1-97rgl_0.92.798 Hmisc_3.8-3survival_2.36-2
ggplot2_0.8.9 proto_0.3-8
[31] reshape_0.8.3 plyr_1.4 foreign_0.8-41
loaded via a namespace (and not attached):
[1] cluster_1.13.2 digest_0.4.2 Matrix_0.999375-46 tools_2.11.1
As i described in the original post I ran
example(ctree)
and the following part yields an error:
if (require("ipred")) {
data("GBSG2", package = "ipred")
GBSG2ct <- ctree(Surv(time, cens) ~ .,data = GBSG2)
plot(GBSG2ct)
treeresponse(GBSG2ct, newdata = GBSG2[1:2,])
}
Error in Summary.Surv(c(1814, 2018, 712, 1807, 772, 448, 2172, 2161, 471, :
Invalid operation on a survival time
//M
On 28. apr. 2011, at 14.19, Jonathan Daily wrote:
It would help people who know more about R's guts than me if you
posted your sessionInfo() output and exactly what commands produced
your error.
It is also recommended that you try simply upgrading R to the latest
version and see if you get an error with the latest version of
'party'. My guess is that the error will go away.
On Wed, Apr 27, 2011 at 3:40 PM, moleps wrote:
Forgot to mention that the ctree command is from the party library.
//M
__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
--
===
Jon Daily
Technician
===
#!/usr/bin/env outside
# It's great, trust me.
__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
Re: [R] ctree and survival problem
sessionInfo yields the following:
> sessionInfo()
R version 2.11.1 (2010-05-31)
x86_64-apple-darwin9.8.0
locale:
[1] en_US.UTF-8/en_US.UTF-8/C/C/en_US.UTF-8/en_US.UTF-8
attached base packages:
[1] stats4tcltk splines grid stats graphics grDevices
utils datasets methods base
other attached packages:
[1] ipred_0.8-8class_7.3-3mlbench_2.1-0 rpart_3.1-46
party_0.9-1vcd_1.2-9 colorspace_1.0-1 strucchange_1.4-3
coin_1.0-18mvtnorm_0.9-95
[11] modeltools_0.2-17 Design_2.3-0 car_2.0-8 nnet_7.3-1
sandwich_2.2-6 boot_1.2-43Amelia_1.2-18 lmtest_0.9-27
zoo_1.6-4 MASS_7.3-7
[21] mgcv_1.7-2 cem_1.0.142randomForest_4.6-2 lattice_0.19-17
nlme_3.1-97rgl_0.92.798 Hmisc_3.8-3survival_2.36-2
ggplot2_0.8.9 proto_0.3-8
[31] reshape_0.8.3 plyr_1.4 foreign_0.8-41
loaded via a namespace (and not attached):
[1] cluster_1.13.2 digest_0.4.2 Matrix_0.999375-46 tools_2.11.1
As i described in the original post I ran
> example(ctree)
and the following part yields an error:
if (require("ipred")) {
data("GBSG2", package = "ipred")
GBSG2ct <- ctree(Surv(time, cens) ~ .,data = GBSG2)
plot(GBSG2ct)
treeresponse(GBSG2ct, newdata = GBSG2[1:2,])
}
Error in Summary.Surv(c(1814, 2018, 712, 1807, 772, 448, 2172, 2161, 471, :
Invalid operation on a survival time
//M
On 28. apr. 2011, at 14.19, Jonathan Daily wrote:
> It would help people who know more about R's guts than me if you
> posted your sessionInfo() output and exactly what commands produced
> your error.
>
> It is also recommended that you try simply upgrading R to the latest
> version and see if you get an error with the latest version of
> 'party'. My guess is that the error will go away.
>
> On Wed, Apr 27, 2011 at 3:40 PM, moleps wrote:
>> Forgot to mention that the ctree command is from the party library.
>> //M
>>
>> __
>> R-help@r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
>> and provide commented, minimal, self-contained, reproducible code.
>>
>
>
>
> --
> ===
> Jon Daily
> Technician
> ===
> #!/usr/bin/env outside
> # It's great, trust me.
__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
Re: [R] ctree and survival problem
It would help people who know more about R's guts than me if you posted your sessionInfo() output and exactly what commands produced your error. It is also recommended that you try simply upgrading R to the latest version and see if you get an error with the latest version of 'party'. My guess is that the error will go away. On Wed, Apr 27, 2011 at 3:40 PM, moleps wrote: > Forgot to mention that the ctree command is from the party library. > //M > > __ > R-help@r-project.org mailing list > https://stat.ethz.ch/mailman/listinfo/r-help > PLEASE do read the posting guide http://www.R-project.org/posting-guide.html > and provide commented, minimal, self-contained, reproducible code. > -- === Jon Daily Technician === #!/usr/bin/env outside # It's great, trust me. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] Ctree Model Variables
On Thu, 24 Mar 2011, jdanielnd wrote:
Hello!
I am not familiar to deal with S4 objects in R, so this question can be
stupid, but I hope I can get an answer. :P
I'm trying to extract what are the response and explanatory variables from a
Binary Tree and Random Forest.
I could already extract the response variable from a Binary Tree using the
response method specified on documentation. But Random Forest didn't had a
similar method.
And regarding the explanatory variables, I couldn't find any solution at
all. It seems to be a simple task to extract the explanatory variables from
a Binary Tree, I just don't know how to deal with the "Slots". In the case
of the Random Forest, the task can be more complicated since it deal with
several trees. What I want is the name of all variables used in any tree.
Does someone have any clue on how should I try to solve this?
ctree() contains a "data" slot which contains an object of
"ModelEnvFormula". This provides the data along with meta-information
which variables are input and response variables. See
help("ModelEnvFormula", package = "modeltools")
for details and examples. As a simple application using ctree(), do the
following: (1) Fit a ctree() named ct. (2) Extract the associated
ModelEnvFormula call mf. (3) Inspect the components...
R> ct <- ctree(Species ~ ., data = iris)
R> mf <- ct@data
R> mf
A ModelEnvFormula with
response variable(s): Species
input variable(s): Sepal.Length Sepal.Width Petal.Length Petal.Width
number of observations: 150
R> head(mf@get("response"))
Species
1 setosa
2 setosa
3 setosa
4 setosa
5 setosa
6 setosa
R> head(mf@get("input"))
Sepal.Length Sepal.Width Petal.Length Petal.Width
1 5.1 3.5 1.4 0.2
2 4.9 3.0 1.4 0.2
3 4.7 3.2 1.3 0.2
4 4.6 3.1 1.5 0.2
5 5.0 3.6 1.4 0.2
6 5.4 3.9 1.7 0.4
Best,
Z
Many thanks!
Joao Daniel Duarte
UFMG - Brazil
--
View this message in context:
http://r.789695.n4.nabble.com/Ctree-Model-Variables-tp3402604p3402604.html
Sent from the R help mailing list archive at Nabble.com.
__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
Re: [R] Ctree Question
On Fri, 9 Jul 2010, steve_fried...@nps.gov wrote: Hello, I've been using ctree and have developed a 55 node - 28 terminal solution. As can be imagined, the plot is difficult to travel down each of the major branches. I've read the help files for ctree I saw where terminal nodes can be color coded. plot(airct, type = "simple") plot(airct, terminal_panel = node_boxplot(airct, col = "blue", + fill = hsv(2/3, 0.5, 1))) Here is my question: Since my model has 55 nodes and 28 terminal nodes,(ie many branches) is it feasible to color code the each of the major branches and track the paths down the decision tree? I'm not sure what exactly you want: Is it to plot Node XY in a certain color, say blue, and also have all branches that lead to that node in blue? If so, you could in principle write a panel function for that but it wouldn't be easy, I'm afraid. Sorry, Z R 2.11.1, Windox XP Thanks Steve Steve Friedman Ph. D. Spatial Statistical Analyst Everglades and Dry Tortugas National Park 950 N Krome Ave (3rd Floor) Homestead, Florida 33034 steve_fried...@nps.gov Office (305) 224 - 4282 Fax (305) 224 - 4147 __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] Ctree Question
Steve, I'm not sure if your task could be accomplished with a ready-made function in party. But, if you could manage to convert your tree structure to a dendrogram, then it's straightforward using dendrapply. In fact, there is an example in dendrapply help page showing how leaves are colored. ?dendrapply HTH. H -Original Message- From: r-help-boun...@r-project.org [mailto:r-help-boun...@r-project.org] On Behalf Of steve_fried...@nps.gov Sent: Friday, July 09, 2010 10:22 AM To: r-help@r-project.org Subject: [R] Ctree Question Hello, I've been using ctree and have developed a 55 node - 28 terminal solution. As can be imagined, the plot is difficult to travel down each of the major branches. I've read the help files for ctree I saw where terminal nodes can be color coded. plot(airct, type = "simple") > plot(airct, terminal_panel = node_boxplot(airct, col = "blue", + fill = hsv(2/3, 0.5, 1))) Here is my question: Since my model has 55 nodes and 28 terminal nodes,(ie many branches) is it feasible to color code the each of the major branches and track the paths down the decision tree? R 2.11.1, Windox XP Thanks Steve Steve Friedman Ph. D. Spatial Statistical Analyst Everglades and Dry Tortugas National Park 950 N Krome Ave (3rd Floor) Homestead, Florida 33034 steve_fried...@nps.gov Office (305) 224 - 4282 Fax (305) 224 - 4147 __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] ctree ordering nodes
On Tue, 6 Jul 2010, Paras Sharma wrote:
Hello,
When using the ctree function, from library (party) what is the syntax to
order the Variables in the nodes in a specific way?
For example, how would I specify to make a binary come first, then a
continuous variable?
Not sure what you mean here. There is only one splitting variable in each
node... The order in which the variables are tested is simply the order in
which they appear in the model formula.
If you want to force ctree() to use a specific variable in the first split
and a specific variable in the second split, this is currently not
possible. ctree() always selects the locally optimal splitting variable.
Also is there a way to force ctree to show variables which are not
significant?
You can inspect the fitted ctree in various ways. See
vignette("party", package = "party")
for a few helpful examples, especially Section 5.3.
If you want to construct your own trees using different algorithms, the
package "partykit" might be of interest which is currently under
development on R-Forge.
hth,
Z
__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
Re: [R] ctree
Joris,
Thank you, your solution did what I needed.
Steve Friedman Ph. D.
Spatial Statistical Analyst
Everglades and Dry Tortugas National Park
950 N Krome Ave (3rd Floor)
Homestead, Florida 33034
steve_fried...@nps.gov
Office (305) 224 - 4282
Fax (305) 224 - 4147
Joris Meys
To
steve_fried...@nps.gov
06/21/2010 10:03 cc
AMr-help@r-project.org
Subject
Re: [R] ctree
You can't get them out of the plot function, but you can calculate
them from the fit. This code returns a matrix with the appropriate
rownames and colnames.
x1 <- treeresponse(iris.ct)
p <- matrix(
unlist(unique(x1)),
ncol=3,
byrow=T
)
colnames(p) <- levels(iris$Species)
rownames(p) <- unique(where(iris.ct))
p
On Mon, Jun 21, 2010 at 3:28 PM, wrote:
>
> Hello,
>
> This is a re-submittal of question I submitted last week, but haven't
rec'd
> any responses.
>
> I need to extract the probabilities used to construct the barplots
> displayed as part of the graph produced by plot("ctree").
>
> For example,
>
> library(party)
>
> iris.ct <- ctree(Species ~ . , data = iris)
> plot(iris.ct)
>
> Instead of a simple example with only 4 terminal nodes, my analysis
> produces 55 terminal nodes, so unless someone is really interested I'm
only
> using the iris data as an example.
>
> This is the model I am sending to ctree. The graph produced is very
> informative, but I need the information from the plot(Marshct)
>
> Marshct <- ctree(Physiogomy ~ meanAnnualDepthAve + meanAnnualDepthWetAve
+
> medAnnualDepthAve + medianAnnualDepthWetAve + medianAnnualDepthDryAve +
> continHydroWetAve + DCHperiodAverage + DCHydroDryAve +
> threeDayWaterDepthMinAve + threeDayWaterDepthMaxAve + sevenDayDryFreqAve
+
> sevenDayWaterDepthMaxAve + sevenDayWaterDepthMinAve +
> seventeenDayWaterDepMaxAve + seventeenDayWaterDepMinAve +
> thirtyoneDayWaterDepthMaxAve + wetIntensityAve + BulkDensity + LOI + TP +
> TN + TC + Total_Mg, data = Marsh)
>
> plot(Marshct)
>
>
> Working in Windows XP with R2.11.1 (2010-05-31)
>
> Thanks
> Steve
>
> Steve Friedman Ph. D.
> Spatial Statistical Analyst
> Everglades and Dry Tortugas National Park
> 950 N Krome Ave (3rd Floor)
> Homestead, Florida 33034
>
> steve_fried...@nps.gov
> Office (305) 224 - 4282
> Fax (305) 224 - 4147
>
> __
> R-help@r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>
--
Joris Meys
Statistical consultant
Ghent University
Faculty of Bioscience Engineering
Department of Applied mathematics, biometrics and process control
tel : +32 9 264 59 87
joris.m...@ugent.be
---
Disclaimer : http://helpdesk.ugent.be/e-maildisclaimer.php
__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
Re: [R] ctree
You can't get them out of the plot function, but you can calculate
them from the fit. This code returns a matrix with the appropriate
rownames and colnames.
x1 <- treeresponse(iris.ct)
p <- matrix(
unlist(unique(x1)),
ncol=3,
byrow=T
)
colnames(p) <- levels(iris$Species)
rownames(p) <- unique(where(iris.ct))
p
On Mon, Jun 21, 2010 at 3:28 PM, wrote:
>
> Hello,
>
> This is a re-submittal of question I submitted last week, but haven't rec'd
> any responses.
>
> I need to extract the probabilities used to construct the barplots
> displayed as part of the graph produced by plot("ctree").
>
> For example,
>
> library(party)
>
> iris.ct <- ctree(Species ~ . , data = iris)
> plot(iris.ct)
>
> Instead of a simple example with only 4 terminal nodes, my analysis
> produces 55 terminal nodes, so unless someone is really interested I'm only
> using the iris data as an example.
>
> This is the model I am sending to ctree. The graph produced is very
> informative, but I need the information from the plot(Marshct)
>
> Marshct <- ctree(Physiogomy ~ meanAnnualDepthAve + meanAnnualDepthWetAve +
> medAnnualDepthAve + medianAnnualDepthWetAve + medianAnnualDepthDryAve +
> continHydroWetAve + DCHperiodAverage + DCHydroDryAve +
> threeDayWaterDepthMinAve + threeDayWaterDepthMaxAve + sevenDayDryFreqAve +
> sevenDayWaterDepthMaxAve + sevenDayWaterDepthMinAve +
> seventeenDayWaterDepMaxAve + seventeenDayWaterDepMinAve +
> thirtyoneDayWaterDepthMaxAve + wetIntensityAve + BulkDensity + LOI + TP +
> TN + TC + Total_Mg, data = Marsh)
>
> plot(Marshct)
>
>
> Working in Windows XP with R2.11.1 (2010-05-31)
>
> Thanks
> Steve
>
> Steve Friedman Ph. D.
> Spatial Statistical Analyst
> Everglades and Dry Tortugas National Park
> 950 N Krome Ave (3rd Floor)
> Homestead, Florida 33034
>
> steve_fried...@nps.gov
> Office (305) 224 - 4282
> Fax (305) 224 - 4147
>
> __
> R-help@r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>
--
Joris Meys
Statistical consultant
Ghent University
Faculty of Bioscience Engineering
Department of Applied mathematics, biometrics and process control
tel : +32 9 264 59 87
joris.m...@ugent.be
---
Disclaimer : http://helpdesk.ugent.be/e-maildisclaimer.php
__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
Re: [R] ctree
On Mon, 21 Jun 2010, steve_fried...@nps.gov wrote:
Hello,
This is a re-submittal of question I submitted last week, but haven't rec'd
any responses.
I need to extract the probabilities used to construct the barplots
displayed as part of the graph produced by plot("ctree").
For example,
library(party)
iris.ct <- ctree(Species ~ . , data = iris)
plot(iris.ct)
You can compute the information from
where(iris.ct)## terminal node ids
treeresponse(iris.ct) ## associated predicted probabilities
So, you could do
tapply(treeresponse(iris.ct), where(iris.ct), unique)
hth,
Z
Instead of a simple example with only 4 terminal nodes, my analysis
produces 55 terminal nodes, so unless someone is really interested I'm only
using the iris data as an example.
This is the model I am sending to ctree. The graph produced is very
informative, but I need the information from the plot(Marshct)
Marshct <- ctree(Physiogomy ~ meanAnnualDepthAve + meanAnnualDepthWetAve +
medAnnualDepthAve + medianAnnualDepthWetAve + medianAnnualDepthDryAve +
continHydroWetAve + DCHperiodAverage + DCHydroDryAve +
threeDayWaterDepthMinAve + threeDayWaterDepthMaxAve + sevenDayDryFreqAve +
sevenDayWaterDepthMaxAve + sevenDayWaterDepthMinAve +
seventeenDayWaterDepMaxAve + seventeenDayWaterDepMinAve +
thirtyoneDayWaterDepthMaxAve + wetIntensityAve + BulkDensity + LOI + TP +
TN + TC + Total_Mg, data = Marsh)
plot(Marshct)
Working in Windows XP with R2.11.1 (2010-05-31)
Thanks
Steve
Steve Friedman Ph. D.
Spatial Statistical Analyst
Everglades and Dry Tortugas National Park
950 N Krome Ave (3rd Floor)
Homestead, Florida 33034
steve_fried...@nps.gov
Office (305) 224 - 4282
Fax (305) 224 - 4147
__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
Re: [R] ctree - party package multivariate response variables
On 2010-03-09 04:40, valeriano.parravic...@unige.it wrote: Hi, I have a problem with ctree of party package. I have data on distribution of more than one species (about 50 species) and I would like identify the relation of this multivariate object (species distribution) with a number of explanatory variables. rs is the name of my dataframe containing the species (columns from 2 to 51) and the explanatory variables (columns 52 and 53). Rows are my sampling sites. I wrote: species<-rs[,2:51] v1<-rs[,52] v2<-rs[53] tree<-ctree(species~v1+v2) It does not work , but when I use the same formula for the univariate case (i.e. a single column - e.g. the total number of species in each samplig sites) it works. I know that ctree can handle multivariate response variables, but I cannot figure out how to do that. The response variables needs to be explicitly specified, e.g. ctree(y1 + y2 ~ x1 + x2) gives you a bivariate response. HTH, Henric Someone can help me? Thank you Valeriano __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] ctree - party package multivariate response variables
Your description of your data isn't clear to me. What are the values in column 2, for example? Are you trying to construct regression or classification trees? A reproducible example would really help. -Peter Ehlers On 2010-03-08 20:40, valeriano.parravic...@unige.it wrote: Hi, I have a problem with ctree of party package. I have data on distribution of more than one species (about 50 species) and I would like identify the relation of this multivariate object (species distribution) with a number of explanatory variables. rs is the name of my dataframe containing the species (columns from 2 to 51) and the explanatory variables (columns 52 and 53). Rows are my sampling sites. I wrote: species<-rs[,2:51] v1<-rs[,52] v2<-rs[53] tree<-ctree(species~v1+v2) It does not work , but when I use the same formula for the univariate case (i.e. a single column - e.g. the total number of species in each samplig sites) it works. I know that ctree can handle multivariate response variables, but I cannot figure out how to do that. Someone can help me? Thank you Valeriano __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. -- Peter Ehlers University of Calgary __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] ctree (party) changing font sizes in plots
On Fri, 20 Nov 2009, Sam Thomas wrote:
When plotting Binary Trees (ctree) from the party package, is there a
way to adjust the font sizes of the leaves?
Currently, this is not implemented in "party". In interactive usage it is
not easy to change the fontsize. If you're printing to PDF, you could
easily increase/decrease the height and width which will decrease/increase
the relative size of the text, respectively. Not elegant but works.
Alternatively, you could obtain the "partykit" package from R-Forge which
has a new implementation of general infrastructure for tree. With that you
can do
library("partykit")
irisct <- ctree(Species ~ ., data = iris)
plot(irisct, gp = gpar(fontsize = 20))
Note that "party" and "partykit" should not be used together, partykit has
a slightly enhanced/modified ctree() function.
hth,
Z
require(party)
irisct <- ctree(Species ~ ., data = iris)
plot(irisct)
I want to adjust the font sizes for "Node 2", "Node 5", etc. I'd also
like to be able to adjust the font sizes for the x-axis and y-axis
labels of the histograms.
Thanks,
Sam Thomas
Revelant Technologies, LLC.
sam.tho...@revelanttech.com
317-696-9214
[[alternative HTML version deleted]]
__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
__
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
Re: [R] ctree (party) plot meaning question
On Mon, 2008-06-30 at 10:41 -0700, Birgitle wrote: > I tried to use ctree but am not sure about the meaning of the plot. > > My.data.ct<-ctree(Resp~., data=My.data) > plot(My.data.ct) > > My data.frame contains 88 explanatory variables (continous,ordered/unordered > multistate,count data) and one response with two groups. > > In the plot are only two variables shown (2 internal nodes) and 3 final > nodes. Does it mean that only these two variables show a significant > asssociation with the response? > > :confused: > > Many thanx in advance Yes, very simply. Nodes are only split if a split has a p-value of less than 1-mincriterion, where mincriterion is 0.95 by default, in a test of independence between the response variable and the predictor. Using an internal data set: mod <- ctree(Species ~ . , data = iris) plot(mod) The plot (on my machine) shows 3 internal nodes resulting in 4 leaves. Petal length and petal width are the two selected variables. The Sepal length and width variables are not selected. Now what happens if we reduce mincriterion? (This is a silly example - you wouldn't want to select a split with a p-value that high): mod1 <- ctree(Species ~ . , data = iris, control = ctree_control(mincriterion = 0.8)) plot(mod1) Now we see that a further split on Petal width has been made, but notice the p-value for this split. So nodes are only split if the null hypothesis of independence between a the response and the predictors cannot be rejected at the given level of significance (1 - mincriterion). This is a different approach to rpart/mvpart, where splitting is based on a few simple stopping rules and then cross-validation is used to prune the tree back. You'd be best to read the cited references in ?ctree for more background on these conditional inference trees. HTH G > > B. > > > > > > - > The art of living is more like wrestling than dancing. > (Marcus Aurelius) __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
