Create frequency tables for multiple factor columns in R

Question

I am a novice in R. I am compiling a separate manual on the syntax for the common functions/features for my work. My sample dataframe as follows:

x.sample <-
structure(list(Q9_A = structure(c(5L, 3L, 5L, 3L, 5L, 3L, 1L, 
5L, 5L, 5L), .Label = c("Impt", "Neutral", "Not Impt at all", 
"Somewhat Impt", "Very Impt"), class = "factor"), Q9_B = structure(c(5L, 
5L, 5L, 3L, 5L, 5L, 3L, 5L, 3L, 3L), .Label = c("Impt", "Neutral", 
"Not Impt at all", "Somewhat Impt", "Very Impt"), class = "factor"), 
Q9_C = structure(c(3L, 5L, 5L, 3L, 5L, 5L, 3L, 5L, 5L, 3L
), .Label = c("Impt", "Neutral", "Not Impt at all", "Somewhat Impt", 
"Very Impt"), class = "factor")), .Names = c("Q9_A", "Q9_B", 
"Q9_C"), row.names = c(NA, 10L), class = "data.frame")

> x.sample
          Q9_A            Q9_B            Q9_C
1        Very Impt       Very Impt Not Impt at all
2  Not Impt at all       Very Impt       Very Impt
3        Very Impt       Very Impt       Very Impt
4  Not Impt at all Not Impt at all Not Impt at all
5        Very Impt       Very Impt       Very Impt
6  Not Impt at all       Very Impt       Very Impt
7             Impt Not Impt at all Not Impt at all
8        Very Impt       Very Impt       Very Impt
9        Very Impt Not Impt at all       Very Impt
10       Very Impt Not Impt at all Not Impt at all

My original dataframe has 21 columns.

If I want to find the mean (treating this as an ordinal variable):

> sapply(x.sample,function(x) mean(as.numeric(x), na.rm=TRUE))
Q9_A Q9_B Q9_C 
 4.0  4.2  4.2

I would like to tabulate a frequency table for ALL the variables in my dataframe. I searched the internet and many forums and saw that the nearest command to do this is using sapply. But when I did it, it gave all 0s.

> sapply(x.sample,function(x) table(factor(x.sample, levels=c("Not Impt at all", "Somewhat Impt",            "Neutral", "Impt", "Very Impt"), ordered=TRUE)))
                Q9_A Q9_B Q9_C
Not Impt at all    0    0    0
Somewhat Impt      0    0    0
Neutral            0    0    0
Impt               0    0    0
Very Impt          0    0    0

QUESTION How can I make use of sapply to tabulate a frequency chart as per the above table for all the columns (that are factors) in a dataframe?

PS So sorry if this seems trivia but I have searched for 2 days without an answer and trying all possible combinations. Maybe I didn't search hard enough =(

Thanks very much.

Answer 1

You were nearly there. Just one small change in your function would have got you there. The x in function(x) ... needs to be passed through to the table() call:

levs <- c("Not Impt at all", "Somewhat Impt", "Neutral", "Impt", "Very Impt")
sapply(x.sample, function(x) table(factor(x, levels=levs, ordered=TRUE)))

A little re-jig of the code might make it a bit easier to read too:

sapply(lapply(x.sample,factor,levels=levs,ordered=TRUE), table)

#                Q9_A Q9_B Q9_C
#Not Impt at all    3    4    4
#Somewhat Impt      0    0    0
#Neutral            0    0    0
#Impt               1    0    0
#Very Impt          6    6    6

Answer 2

Coming a bit late, but here's a reshape2 possible solution. It could have been very straightforward with recast but we need to handle empty factor levels here so we need to specify both factorsAsStrings = FALSE within melt and drop = FALSE within dcast, while recast can't pass arguments to melt (only to dcast), so here goes

library(reshape2)
x.sample$indx <- 1 
dcast(melt(x.sample, "indx", factorsAsStrings = FALSE), value ~ variable, drop = FALSE)
#             value Q9_A Q9_B Q9_C
# 1            Impt    1    0    0
# 2         Neutral    0    0    0
# 3 Not Impt at all    3    4    4
# 4   Somewhat Impt    0    0    0
# 5       Very Impt    6    6    6

If we wouldn't care about empty levels a quick solution would be just

recast(x.sample, value ~ variable, id.var = "indx")
#             value Q9_A Q9_B Q9_C
# 1            Impt    1    0    0
# 2 Not Impt at all    3    4    4
# 3       Very Impt    6    6    6

---

Alternatively, if speed is a concern, we can do the same using data.atble

library(data.table)
dcast(melt(setDT(x.sample), measure.vars = names(x.sample), value.factor = TRUE), 
           value ~ variable, drop = FALSE)
#              value Q9_A Q9_B Q9_C
# 1:            Impt    1    0    0
# 2:         Neutral    0    0    0
# 3: Not Impt at all    3    4    4
# 4:   Somewhat Impt    0    0    0
# 5:       Very Impt    6    6    6