Create a new column by applying a custom function to selected columns only in R-CodePudding

I want to apply a custom function to every column of df and assign the value that function returns to a new column in that dataframe. My function takes a vector of values from chosen columns (in my case values from columns 12:17 will be used), and returns a calculated value (diversity index). The function is defined as:

shannon <- function(p){
  if (0 %in% p) {
    p = replace(p,p==0,0.0001)
  } else {
    p
  }
  H = -sum(p*log(p))
  return (H)
}

A random row from the dataset looks like this:

p <- df[3000,12:17]
        x1        x2        x3        x4         x5 x6
 0.5777778 0.1777778 0.1555556 0.2888889 0.02222222  0

When I apply the custom function to this row, like this:

shannon(as.vector(t(p)))

It returns the correctly calculated value of 1.357692.

Now, I want to make this value into a new column of my dataset, by applying the custom function to the specific columns form my dataset. I try to do it using mutate and sapply by running:

df <- mutate(df, shannon = sapply(as.vector(t(census[,12:17])), shannon))

but it returns

Error in `mutate()`:
! Problem while computing `shannonVal = sapply(as.vector(t(census[, 12:17])), shannon)`.
✖ `shannonVal` must be size 9467 or 1, not 56802.

The number of rows in my dataset is 9467, so the sapply is returning something that's 6 times as long. But why, and how can I fix it?

CodePudding user response：

Building on Ric's comment, df <- mutate(df, shannon = apply(census[,12:17], 1, function(x) {shannon(t(x)}) might just do the trick

CodePudding user response：

Ric's answer works: df$shannon <- apply(df[,12:17], 1, shannon)

df and census are the same thing, sorry for the confusion