Counting the number of factors for several variables and summing the results in one table
This is my first post here and I am very new to programming and R. so please excuse any nonsense.
I have the following framework:
a <- data.frame("sickness1" = c(1,1,2,3,3,5,6, 4, 4, 4),
"sickness2" = c(NA, NA, 3, 3, 4, 6, 1, 2, 5, 6),
"sickness3" = c(NA, NA, 3, 4, 4, 6, 1, 2, 5, 6),
"sickness4" = c(NA, NA, 6, 3, 4, 6, 1, 2, 5, 6))
each line represents one case. each column is an ordered factor variable. I converted the variables to factors like this (using a hint I found on stackoverflow!):
a[] <- lapply(a, factor,
levels = c(1:6),
labels = c(3, 25, 50, 75, 97, 100))
I would like to receive the following output:
percent sickness1 sickness2 sickness3 sickness4
1 3 1 1 1 2
2 25 1 1 1 1
3 50 2 1 1 2
4 75 1 2 1 3
5 97 1 1 1 1
6 100 2 2 3 1
I have already found a solution which is for a very long time:
# counting
ab <- ldply(lapply(a, count))
#getting it into the right format
ab2 <- dcast(
data = ab,
formula = x ~ .id,
value.var = "freq")
# changing the name of the first column
colnames(ab2)[1] <- "percent"
#deleting row 7 cause it contains the NAs which I dont want to have
ab2 <- ab2[-7,]
ab2
is there a faster and easier way to do this? just somehow use ddply? the result that summary (a) gives is too confusing and I don't know how I could manipulate it to look the way I want it to. Also, the real data I'm working with is much larger and I have to do this many times ....
source to share
ok, so I found that two options are possible:
Nr1 by akrun:
un1 <- as.character(sort(unique(unlist(a, use.names=FALSE))))
data.frame(percent=un1,do.call(cbind,
lapply(a, function(x) table(factor(x, levels=un1)))))
Nr.2: alexis_laz:
Considering that I could easily make the data look like this: (this is the dataframe just above with a column added for institution)
a <- data.frame("institution" = c(1:10), "sickness1" = c(1,1,2,3,3,5,6, 4, 4, 4),
"sickness2" = c(NA, NA, 3, 3, 4, 6, 1, 2, 5, 6),
"sickness3" = c(NA, NA, 3, 4, 4, 6, 1, 2, 5, 6),
"sickness4" = c(NA, NA, 6, 3, 4, 6, 1, 2, 5, 6))
a[-1] <- lapply(a[-1], factor,
levels = c(1:6),
labels = c("0 to 3%","4-25%", "25-50%", "51-75%","76-97%","97-100%"))
Then I could convert this wide data form to long data format like this:
b2 <- melt(a, id.vars = "institution")
then the usual table function works:
table(b2[[3]], b2[[2]])
note that order matters
Thanks a lot guys!
source to share
Basically, this is a variant of the answer to the topic. Use stack
and table
together, for example:
as.data.frame.matrix( ## converts the output to a data.frame
table( ## does the actual tabulation
stack( ## stack makes your data.frame long
lapply(a, as.character)), ## but won't work with factors; convert to char
useNA = "no") ## we don't want NA values
)[levels(a[[1]]), ] ## We want our rows in a nicer order
# sickness1 sickness3 sickness4 sickness5
# 3 2 1 1 1
# 25 1 1 1 1
# 50 2 2 1 1
# 75 3 1 2 1
# 97 1 1 1 1
# 100 1 2 2 3
Alternatively, the "dplyr" + "tidyr" approach is used here:
library(dplyr)
library(tidyr)
a %>% gather(var, val, sickness1:sickness5) %>% ## make the data long
mutate(val = factor(val, levels(unlist(a)))) %>% ## refactor "val" column
rev %>% ## reverse the order of val and var
table %>% ## make your table
as.data.frame.matrix ## convert it to a data.frame
# sickness1 sickness3 sickness4 sickness5
# 3 2 1 1 1
# 25 1 1 1 1
# 50 2 2 1 1
# 75 3 1 2 1
# 97 1 1 1 1
# 100 1 2 2 3
source to share