In SPSS it is fairly easy to create a summary table of categorical variables using "Custom Tables":
How can I do this in R?
General and expandable solutions are preferred, and solutions using the Plyr and/or Reshape2 packages, because I am trying to learn those.
Example Data: (mtcars is in the R installation)
df <- colwise(function(x) as.factor(x) ) (mtcars[,8:11])
P.S.
Please note, my goal is to get everything in one table like in the picture. I have been strugling for many hours but my attempts have been so poor that posting the code probably won't add to the comprehensibility of the question.
One way to get the output, but not the formatting:
library(plyr)
ldply(mtcars[,8:11],function(x) t(rbind(names(table(x)),table(x),paste0(prop.table(table(x))*100,"%"))))
.id 1 2 3
1 vs 0 18 56.25%
2 vs 1 14 43.75%
3 am 0 19 59.375%
4 am 1 13 40.625%
5 gear 3 15 46.875%
6 gear 4 12 37.5%
7 gear 5 5 15.625%
8 carb 1 7 21.875%
9 carb 2 10 31.25%
10 carb 3 3 9.375%
11 carb 4 10 31.25%
12 carb 6 1 3.125%
13 carb 8 1 3.125%
A base R solution using lapply() and do.call() with rbind() to stitch together the pieces:
x <- lapply(mtcars[, c("vs", "am", "gear", "carb")], table)
neat.table <- function(x, name){
xx <- data.frame(x)
names(xx) <- c("Value", "Count")
xx$Fraction <- with(xx, Count/sum(Count))
data.frame(Variable = name, xx)
}
do.call(rbind, lapply(seq_along(x), function(i)neat.table(x[i], names(x[i]))))
Results in:
Variable Value Count Fraction
1 vs 0 18 0.56250
2 vs 1 14 0.43750
3 am 0 19 0.59375
4 am 1 13 0.40625
5 gear 3 15 0.46875
6 gear 4 12 0.37500
7 gear 5 5 0.15625
8 carb 1 7 0.21875
9 carb 2 10 0.31250
10 carb 3 3 0.09375
11 carb 4 10 0.31250
12 carb 6 1 0.03125
13 carb 8 1 0.03125
The rest is formatting.
Here's my solution. It ain't pretty, which is why I put a bag over its head (wrap it in a function). I also add another variable to demonstrate that it's general (I hope).
prettyTable <- function(x) {
tbl <- apply(x, 2, function(m) {
marc <- sort(unique(m))
cnt <- matrix(table(m), ncol = 1)
out <- cbind(marc, cnt)
out <- out[order(marc), ] # do sorting
out <- cbind(out, round(prop.table(out, 2)[, 2] * 100, 2))
})
x2 <- do.call("rbind", tbl)
spaces <- unlist(lapply(apply(x, 2, unique), length))
space.names <- names(spaces)
spc <- rep("", sum(spaces))
ind <- cumsum(spaces)
ind <- abs(spaces - ind)+1
spc[ind] <- space.names
out <- cbind(spc, x2)
out <- as.data.frame(out)
names(out) <- c("Variable", "Levels", "Count", "Column N %")
out
}
prettyTable(x = mtcars[, c(2, 8:11)])
Variable Levels Count Column N %
1 cyl 4 11 34.38
2 6 7 21.88
3 8 14 43.75
4 vs 0 18 56.25
5 1 14 43.75
6 am 0 19 59.38
7 1 13 40.62
8 gear 3 15 46.88
9 4 12 37.5
10 5 5 15.62
11 carb 1 7 21.88
12 2 10 31.25
13 3 3 9.38
14 4 10 31.25
15 6 1 3.12
16 8 1 3.12
Using googleVis package, you can make a handy html table.
plot(gvisTable(prettyTable(x = mtcars[, c(2, 8:11)])))
You may find the following code snippet useful. It utilizes the base package functions table, margin.table, and prop.table and does not require any other packages. It does collect the results to a list with named dimensions however (these could be collected to a single matrix with rbind):
dat <- table(mtcars[,8:11])
result <- list()
for(m in 1:length(dim(dat))){
martab <- margin.table(dat, margin=m)
result[[m]] <- cbind(Freq=martab, Prop=prop.table(martab))
}
names(result) <- names(dimnames(dat))
> result
$vs
Freq Prop
0 18 0.5625
1 14 0.4375
$am
Freq Prop
0 19 0.59375
1 13 0.40625
$gear
Freq Prop
3 15 0.46875
4 12 0.37500
5 5 0.15625
$carb
Freq Prop
1 7 0.21875
2 10 0.31250
3 3 0.09375
4 10 0.31250
6 1 0.03125
8 1 0.03125
Here is a solution using the freq function of the questionr package (shameless autopromotion, sorry) :
R> lapply(df, freq)
$vs
n %
0 18 56.2
1 14 43.8
NA 0 0.0
$am
n %
0 19 59.4
1 13 40.6
NA 0 0.0
$gear
n %
3 15 46.9
4 12 37.5
5 5 15.6
NA 0 0.0
$carb
n %
1 7 21.9
2 10 31.2
3 3 9.4
4 10 31.2
6 1 3.1
8 1 3.1
NA 0 0.0
Unfortunately there seems to be no R package yet that can generate a nice output like SPSS. Most functions for generating tables seem to define their own special formats what gets you into trouble if you want to export or work on it in another way.
But I'm sure R is capable of that and so I started writing my own functions. I'm happy to share the result (work in progress-status, but gets the job done) with you:
The following function returns for all factor variables in a data.frame the frequency or the percentage (calc="perc") for each level of the factor variable "variable".
The most important thing may be that the output is a simple & user friendly data.frame. So, compared to many other functions, it's no problem to export the results an work with it in any way you want.
I realize that there is much potential for further improvements, i.e. add a possibility for selecting row vs. column percentage calculation, etc.
contitable <- function( survey_data, variable, calc="freq" ){
# Check which variables are not given as factor
# and exlude them from the given data.frame
survey_data_factor_test <- as.logical( sapply( Survey, FUN=is.factor) )
survey_data <- subset( survey_data, select=which( survey_data_factor_test ) )
# Inform the user about deleted variables
# is that proper use of printing to console during a function call??
# for now it worksjust fine...
flush.console()
writeLines( paste( "\n ", sum( !survey_data_factor_test, na.rm=TRUE),
"non-factor variable(s) were excluded\n" ) )
variable_levels <- levels(survey_data[ , variable ])
variable_levels_length <- length( variable_levels )
# Initializing the data.frame which will gather the results
result <- data.frame( "Variable", "Levels", t(rep( 1, each=variable_levels_length ) ) )
result_column_names <- paste( variable, variable_levels, sep="." )
names(result) <- c("Variable", "Levels", result_column_names )
for(column in 1:length( names(survey_data) ) ){
column_levels_length <- length( levels( survey_data[ , column ] ) )
result_block <- as.data.frame( rep( names(survey_data)[column], each=column_levels_length ) )
result_block <- cbind( result_block, as.data.frame( levels( survey_data[,column] ) ) )
names(result_block) <- c( "Variable", "Levels" )
results <- table( survey_data[ , column ], survey_data[ , variable ] )
if( calc=="perc" ){
results <- apply( results, MARGIN=2, FUN=function(x){ x/sum(x) })
results <- round( results*100, 1 )
}
results <- unclass(results)
results <- as.data.frame( results )
names( results ) <- result_column_names
rownames(results) <- NULL
result_block <- cbind( result_block, results)
result <- rbind( result, result_block )
}
result <- result[-1,]
return( result )
}
来源：https://stackoverflow.com/questions/15111629/create-summary-table-of-categorical-variables-of-different-lengths