Compute median per column in loop

Question

I have this loop to compute the mean per column, which works.

for (i in 1:length(DF1)) {     
    tempA <- DF1[i]                                 # save column of DF1 onto temp variable 
    names(tempA) <- 'word'                          # label temp variable for inner_join function
    DF2 <- inner_join(tempA, DF0, by='word')        # match words with numeric value from look-up DF0
    tempB <- as.data.frame(t(colMeans(DF2[-1])))    # compute mean of column
    DF3<- rbind(tempB, DF3)                         # save results togther
}

The script uses the dplyr package for inner_join.

DF0 is the look-up database with 3 columns (word, value1, value2, value3).
DF 1 is the text data with one word per cell.
DF3 is the output.

Now I want to compute the median instead of the mean. It seemed easy enough with the colMedians function from 'robustbase', but I can't get the below to work.

library(robustbase)

for (i in 1:length(DF1)) {     
    tempA <- DF1[i]
    names(tempA) <- 'word'
    DF2 <- inner_join(tempA, DF0, by='word')
    tempB <- as.data.frame(t(colMedians(DF2[-1])))
    DF3<- rbind(tempB, DF3) 
}

The error message reads:

Error in colMedians(tog[-1]) : Argument 'x' must be a matrix.

I've tried to format DF2 as a matrix prior to the colMedians function, but still get the error message:

Error in colMedians(tog[-1]) : Argument 'x' must be a matrix.

I don't understand what is going on here. Thanks for the help!

Happy to provide sample data and error traceback, but trying to keep it as crisp and simple as possible.

When asking for help, you should include a simple [reproducible example](https://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example) with sample input and desired output that can be used to test and verify possible solutions. — MrFlick, May 21 '18 at 15:21
Have you tried to use the median function from the stats package? — Dave Rosenman, May 21 '18 at 15:22
Change colMedians for apply: tempB <- as.data.frame(apply(DF2[-1], 2, median)) — Juan Antonio Roldán Díaz, May 21 '18 at 15:38
Why `*_all*` approach wouldn't work here as in: `mtcars %>% summarise_all(funs(median))`? I reckon that some sample data would help. — Konrad, May 21 '18 at 16:04
@MrFlick anticipating this comment I said I'd provide data if required :-) guess that is redundant now. — Simone, May 22 '18 at 15:53
@RuiBarradas in the nerd spirit had to try your solution as well. worked perfectly. Want to post it so I can accept it as answer? — Simone, May 22 '18 at 16:02
@JuanAntonioRoldánDíaz the apply family did help. However, I don't need to get rid of the first row, instead need to specify the columns as the joined DF2 inherits the data structure from the look-up DF0. — Simone, May 22 '18 at 16:07

score 2 · Accepted Answer · answered May 22 '18 at 16:17

According to the comment by the OP, the following solved the problem.
I have added a call to library(dplyr).
My contribution was colMedians(data.matrix(DF2[-1]), na.rm = TRUE).

library(robustbase)
library(dplyr)

for (i in 1:length(DF1)) {     
    tempA <- DF1[i]
    names(tempA) <- 'word'
    DF2 <- inner_join(tempA, DF0, by='word')
    tempB <- colMedians(data.matrix(DF2[-1]), na.rm = TRUE)
    DF3 <- rbind(tempB, DF3) 
}

Simone · Answer 2 · 2018-05-22T16:36:28.603

Stumbled on this answer which helped me fix the loop as following:

DF3Mean <- data.frame()                         # instantiate dataframe 
DF4Median <- data.frame(                        # instantiate dataframe

for (i in 1:length(DF1)) {     
tempA <- DF1[i]                                 # save column of DF1 onto temp variable 
names(tempA) <- 'word'                          # label temp variable for inner_join function
DF2 <- inner_join(tempA, DF0, by='word')        # match words with numeric value from look-up DF0
tempMean <- as.data.frame(t(colMeans(DF2[-1]))) # compute mean of column
DF3Mean <- rbind(tempMean, DF3Mean)             # save results togther
tempMedian <- apply(DF2[ ,2:4], 2, median)      #compute mean for columns 2,3, and 4 
DF4Median <- rbind(tempMedian, DF4Median)       # save results togther
}

I guess I was too stuck in my mind on the colMedian function.

Compute median per column in loop

2 Answers2