How to find outliers with non-linear threshold

Question

Assuming that I have a dataframe such as

x <- round(runif(1000,-5,5), 2)
y <- round(runif(1000,0,5), 2)
z <- sprintf("%s%05d", "A", seq.int(1000))
df <- data.frame(x, y, z)

How can I find which data point (names of the point from column z) is an outlier of a non-linear threshold that looks like this

y = a/(|x|-c)

where a and c are values that I can arbitrary chose?

|x| is the modulus of x

Something like `a <- 1; c <- 1; df[df$y > a / (abs(df$x) - c), "z"];`? You could wrap the `a / abs()...` in a function if you wanted for more readability — Mike H., Jan 03 '18 at 15:00

score 1 · Accepted Answer · answered Jan 03 '18 at 18:33

1

As mentioned in the comment, you can create a short function for this:

find_outliers = function(df, a, c){
  y_threshold = a/(abs(df$x)-c)
  return(df$z[df$y>y_threshold])
}

a=1
c=0.1
find_outliers(df,a,c)

answered Jan 03 '18 at 18:33

Kelli-Jean

1 Answers1