Reordering columns in an R dataframe according to average values

Question

I would like to reorder a data frame's columns according to arithmetic mean of each column.

For:

S1 S2 S3
1  1  1
2  1  1
3  3  1 

the expected output is:

S3 S2 S1
1  1  1 
1  1  2 
1  3  3 

In above case, the averages were: S1 = 2, S2 = 1.6666 and S3 = 1, inverting S1 and S3 columns positions in a data frame.

Additionally, my real data have NA´s values too.

Answer 1

Use the order() function.

An exemplary data frame:

df <- data.frame(s1=runif(5), s2=runif(5), s3=runif(5))
df[1,2] <- NA # some NAs
df
##           s1        s2         s3
## 1 0.74473576        NA 0.71547379
## 2 0.66997782 0.6474405 0.62320795
## 3 0.05361586 0.5370381 0.03298139
## 4 0.06209263 0.9409920 0.46096984
## 5 0.42432948 0.9983042 0.38503196

Calculate column means, with NAs omitted:

(mns <- colMeans(df, na.rm=TRUE))
##        s1        s2        s3 
## 0.3909503 0.7809437 0.4435330 

The desired column ordering is:

order(mns)
## [1] 1 3 2

(s1 goes first, s2 goes last, and s3 should become the 2nd column)

Now you may reorder the columns:

(df <- df[,order(mns)])
##           s1         s3        s2
## 1 0.74473576 0.71547379        NA
## 2 0.66997782 0.62320795 0.6474405
## 3 0.05361586 0.03298139 0.5370381
## 4 0.06209263 0.46096984 0.9409920
## 5 0.42432948 0.38503196 0.9983042