Learn R Programming

pamr (version 1.57)

pamr.makeclasses: A function to interactively define classes from a clustering tree

Description

function to interactively define classes from a clustering tree

Usage

pamr.makeclasses(data, sort.by.class = FALSE, ...)

Value

A vector of class labels 1,2,3... If a component is NA (missing), then the sample is not assigned to any class. This vector should be assigned to the newy component of data, for use in pamr.train etc. Note that pamr.train uses the class labels in the component newy'' if it is present. Otherwise it uses the data labels y''.

Arguments

data

The input data. A list with components: x- an expression genes in the rows, samples in the columns, and y- a vector of the class labels for each sample, and batchlabels- a vector of batch labels for each sample. This object if the same form as that produced by pamr.from.excel.

sort.by.class

Optional argument. If true, the clustering tree is forced to put all samples in the same class (as defined by the class labels y in `data') together in the tree. This is useful if a regrouping of classes is desired. Eg: given classes 1,2,3,4 you want to define new classes (1,3) vs (2,4) or 2 vs (1,3)

...

Any additional arguments to be passed to hclust

Author

Trevor Hastie, Robert Tibshirani, Balasubramanian Narasimhan, and Gilbert Chu

Details

pamr.makeclasses Using this function the user interactively defines a new set of classes, to be used in pamr.train, pamr.cv etc. After invoking pamr.makeclasses, a clustering tree is drawn. This callss the R function hclust, and any arguments for hclust can be passed to it. Using the left button, the user clicks at the junction point defining the subgroup 1. More groups can be added to class 1 by clicking on further junction points. The user ends the definition of class 1 by clicking on the rightmost button (in Windows, an additional menu appears and he chooses Stop). This process is continued for classes 2,3 etc. Note that some sample may be left out of the new classes. Two consecutive clicks of the right button ends the definition for all classes.

At the end, the clustering is redrawn, with the new class labels shown.

Note: this function is "fragile". The user must click close to the junction point, to avoid confusion with other junction points. Classes 1,2,3.. cannot have samples in common (if they do, an Error message will appear). If the function is confused about the desired choices, it will complain and ask the user to rerun pamr.makeclasses. The user should also check that the labels on the final redrawn cluster tree agrees with the desired classes.

Examples

Run this code

suppressWarnings(RNGversion("3.5.0"))
set.seed(120)
#generate some data
x <- matrix(rnorm(1000*40),ncol=40)
y <- sample(c(1:4),size=40,replace=TRUE)
batchlabels <- sample(c(1:5),size=40,replace=TRUE)

mydata <- list(x=x,y=factor(y),batchlabels=factor(batchlabels),
               geneid=as.character(1:nrow(x)),
               genenames=paste("g",as.character(1:nrow(x)),sep=""))

# mydata$newy <- pamr.makeclasses(mydata) Run this and define some new classes

train <- pamr.train(mydata)

Run the code above in your browser using DataLab