Scatterplots in R using lattice and cloud, how to determine colors by factors?

Question

I am still struggling with R plots and colors -- some results are as I expected, some not.

I have a 2-million point data set, generated by a simulation process. There are several variables on the dataset, but I am interested on three and on a factor that describe the class for that data point.

Here is a short snippet of code that reads the points and get some basic statistics on it:

library(lattice)
library(plyr)
myData <- read.table("dados - b1000 n10000 var 0,2 - MAX40.txt",
                col.names=c("Class","Thet1Thet2","Thet3Thet2","Thet3Thet1",
                            "K12","K23","delta","w_1","w_2","w_3"))
count (myData$Class)

That gives me

##   x    freq
## 1 A    8030
## 2 B   17247
## 3 C    4999
## 4 D   16495
## 5 E 1949884
## 6 N    3345

(the input file is quite large, cannot add it as a link)

I want to see these points in a scatterplot matrix, so I use the code

colors=c("red","green","blue","cyan","magenta","yellow")
# Let's try with a very small dot size, see if we can visualize the inners of the cube.
cloud(myData$delta ~ myData$K12 + myData$K23, xlab="K12", ylab="K23", zlab="delta",
  cex=0.001,main="All Classes",col.point = colors[myData$Class])

Here is the result. As expected, points from class E are in vast majority, so I cannot see points of other classes. The problem is that I expected the points to be plotted in magenta (classes are A, B, C, D, E, N; colors are red, green, blue, cyan, magenta, yellow).

When I do the plot class by class it works as expected, see two examples:

data <- subset(myData, Class=="A")
cloud(data$delta ~ data$K12 + data$K23, xlab="K12", ylab="K23", zlab="delta",pch=20,main="Class A",
      col.point = colors[data$Class])

gives this:

And this snippet of code

data <- subset(myData, Class=="E")
cloud(data$delta ~ data$K12 + data$K23, xlab="K12", ylab="K23", zlab="delta",pch=20,main="Class E",
      col.point = colors[data$Class])

gives this:

This also seems as expected: a plot of points of all classes except E.

data <- subset(myData, Class!="E")
cloud(data$delta ~ data$K12 + data$K23, xlab="K12", ylab="K23", zlab="delta",pch=20,
      cex=0.01,main="All Classes (except E)",col.point = colors[data$Class])

The question is, why on the first plot the points are blue instead of magenta?

This question is somehow similar to Color gradient for elevation data in a XYZ plot with R and Lattice but now I am using factors to determine colors on the scatterplot.

I've also read Changing default colours of a lattice plot by factor -- grouping plots by a factor (using the parameter groups.factor=myData$Class) does not solve my problem, plots are still in blue but separated by class.

Edited to add more information: this fake data set can be used for tests.

num <- 10
data <- as.data.frame(
    cbind(
      x=rep(seq(1,num), each=num*num), 
      y=rep(seq(1,num), each=num), 
      z=rep(seq(1,num)) 
    ))
# This is ugly but works!
data$Class[data$z==1]<-'A'
data$Class[data$z==2]<-'A'
data$Class[data$z==3]<-'B'
data$Class[data$z==4]<-'B'
data$Class[data$z==5]<-'C'
data$Class[data$z==6]<-'C'
data$Class[data$z==7]<-'D'
data$Class[data$z==8]<-'D'
data$Class[data$z==9]<-'E'
data$Class[data$z==10]<-'E'
str(data)

When I plot it with

colors=c("red","green","blue","cyan","magenta","yellow")
cloud(data$z ~ data$x + data$y, xlab="X", ylab="Y", zlab="Z",main="All Classes",
col.point = colors[data$Class])

I get the plot below. All points are in blue.

See [how to create a reproducible example](http://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example). If your real data it too large, use a built-in data set or create some fake data to make your examples reproducible. It's much easier to help when we can run the code. — MrFlick, Oct 07 '15 at 14:48
what is the output of `levels(myData$Class)`. Can you add some data - `dput(head(myData))` — jeremycg, Oct 07 '15 at 14:48
@MrFlick, I've added code to create a compatible fake data set, which can be used to reproduce my problem. — Rafael Santos, Oct 07 '15 at 15:08
@jeremycg: levels(myData$Class) says "A" "B" "C" "D" "E" "N" — Rafael Santos, Oct 07 '15 at 15:19
`col.point` isn't the correct argument, try `col`:`cloud(data$z ~ data$x + data$y, xlab="X", ylab="Y", zlab="Z",main="All Classes", col = colors[data$Class])` — jeremycg, Oct 07 '15 at 15:27
@jeremycg, using the fake data set, I've tried this: cloud(data$z ~ data$x + data$y, xlab="X", ylab="Y", zlab="Z",main="All Classes",col = colors[data$Class]) -- the plot came empty! Please also notice that col.point worked in some examples above. — Rafael Santos, Oct 07 '15 at 15:36
you will need to run: `data$Class <- as.factor(data$Class)` first, apologies. It should work fine on your full data however — jeremycg, Oct 07 '15 at 15:37

Rafael Santos · Answer 1 · 2015-10-07T16:05:53.437

JeremyCG found the problem. Here is the complete code that works, for future reference.

library(lattice)
num <- 10
data <- as.data.frame(
cbind(
  x=rep(seq(1,num), each=num*num), 
  y=rep(seq(1,num), each=num), 
  z=rep(seq(1,num)) 
))
data$Class[data$z==1]<-'A'
data$Class[data$z==2]<-'A'
data$Class[data$z==3]<-'B'
data$Class[data$z==4]<-'B'
data$Class[data$z==5]<-'C'
data$Class[data$z==6]<-'C'
data$Class[data$z==7]<-'D'
data$Class[data$z==8]<-'D'
data$Class[data$z==9]<-'E'
data$Class[data$z==10]<-'E'
str(data)

That showed the issue:

## 'data.frame':    1000 obs. of  4 variables:
##  $ x    : int  1 1 1 1 1 1 1 1 1 1 ...
##  $ y    : int  1 1 1 1 1 1 1 1 1 1 ...
##  $ z    : int  1 2 3 4 5 6 7 8 9 10 ...
##  $ Class: chr  "A" "A" "B" "B" ...

Class must be a factor. This solved it:

data$Class <- as.factor(data$Class)
str(data)

## 'data.frame':    1000 obs. of  4 variables:
##  $ x    : int  1 1 1 1 1 1 1 1 1 1 ...
##  $ y    : int  1 1 1 1 1 1 1 1 1 1 ...
##  $ z    : int  1 2 3 4 5 6 7 8 9 10 ...
##  $ Class: Factor w/ 5 levels "A","B","C","D",..: 1 1 2 2 3 3 4 4 5 5 ...

Then plot it:

colors=c("red","green","blue","cyan","magenta","yellow")
cloud(data$z ~ data$x + data$y, xlab="X", ylab="Y", zlab="Z",
pch=20,main="All Classes",col = colors[data$Class])

Here is the result:

Thanks @jeremycg !

Scatterplots in R using lattice and cloud, how to determine colors by factors?

Edited to add more information: this fake data set can be used for tests.

1 Answers1