Identifying outlier when using a 3d plot in R

Question

I am new to R and have 1024 rows of data with 3 columns of numerical data. I have created a plot3d and I need to identify the row number of an outlier which stands out like a sore thumb in plot3D but in any other graphs is not visable.

Answer 1

Hopefully this helps get the job done for you.

> data <- c(-1.5454, -0.6855, 0.1003, -0.5284, -0.4065, -0.2645, 
            -1.0868, -0.5329, 0.1623, -1e-04, -0.9569, -2.0055, 
            0.389, -0.8356, -2.2085, 0.5326, 0.0391, -0.5044, 
            -1.8376, -0.7834, 0.3436)
## original data
> dd <- data.frame(matrix(data, ncol = 3, byrow = TRUE))
## find the row number of the largest row maximum
> which.max(apply(dd, 1, max))
[1] 6
## Use the previous line to remove the unwanted row
> newDd <- dd[ -which.max(apply(dd, 1, max)), ] 
## plot the two data frames together to see the difference
> library(plot3D)
> par(mfrow = c(1, 2))
> with(dd, scatter3D(X1, X2, X3, phi = 0, theta = 50, bty = "g",
                     col = gg.col(100), pch = 19, cex = 2, colkey = FALSE))
> with(newDd, scatter3D(X1, X2, X3, phi = 0, theta = 50, bty = "g",
                        col = gg.col(100), pch = 19, cex = 2, colkey = TRUE))

Answer 2

Use the built-int arrayInd to find the maximum value (or minimum):

arrayInd(which.max(as.matrix(df)), .dim = dim(df))

For example, we are going to make a 3 column dataframe with one sore thumb.

df <- data.frame(structure(replicate(3, runif(1024, 0, 1), simplify = FALSE), .Names = c('one', 'two', 'three')))
df[50, 2] <- 10

Now we get

arrayInd(which.max(as.matrix(df)), .dim = dim(df))
#      [,1] [,2]
# [1,]   50    2

And we see the offender is in row 50 and column 2.

Answer 3

Welcome user3479729. Please post a reproducible example. Otherwise you will either get no answer or bad ones.

if 'M' is the matrix that you plot and 'thres' is your threshold for outlier data (I need here to assume that you are plotting a matrix?), you can use:

> which(M>thres,arr.ind=TRUE)