通过单个变量对多个变量的数据进行分组

Question

In the following dataset, I want to do two things

pt_id <- c(1,1,1,1,1,2,2,2,3,3,3,3,3,4,4,4,4)
Tobacco <- c("once","twice","never", NA, NA, NA, NA, NA,"Once","Twice","Quit","Once",NA,NA,"Never", NA, "Never")
Alcohol <- c("twice", "once",NA, NA, "never", NA, NA, "Once", NA, "Quit", "Twice", NA, "Once", NA, NA, "Never", "Never")
PA <- c("once",NA,"never", NA, NA, NA, NA, NA,"Once",NA,"Quit","Once",NA,NA,"Never", NA, NA)
mydata <- data.frame(pt_id, Tobacco, Alcohol, PA)
mydata

1. Count the number of rows per patient that are not NA for each variable (Tobacco, alcohol and PA) grouped by patient ID.

I used the following code to get my output but I can do it only for one variable at a time.

mydata_tob <- mydata %>% 
  filter(!is.na(Tobacco)) %>% 
  group_by(pt_id) %>% 
  count()

# A tibble: 3 x 2
# Groups:   pt_id [3]
  pt_id     n
  <dbl> <int>
1     1     3
2     3     4
3     4     2

But this is very time-consuming for me as I have many many variables in my original dataset. I want a similar kind of output for all the variables in one go.

2. My end result is I want to calculate the percentage of pt_id with more than 1 entry for each variable. I created the following function (only for tobacco) to be able to do so

gt1_prop <- function(n) {
  gt1_len <- length(mydata_tob$n[mydata_tob$n > 1])
  len_tot <- length(mydata_tob$n)
  gt1_prop <- (gt1_len/ len_tot)*100
  return(gt1_prop)
}

Again I want to code in a way that I get the proportion for each variable (Tobacco, Alcohol and PA) in the dataset.

Any suggestions will be helpful. Thanks in advance!

Answer 1

To count number of non-NA values for each pt_id you can use across .

library(dplyr)

mydata %>%
  group_by(pt_id) %>%
  summarise(across(Tobacco:PA, ~sum(!is.na(.)))) -> result
result

#  pt_id Tobacco Alcohol    PA
#  <dbl>   <int>   <int> <int>
#1     1       3       3     2
#2     2       0       1     0
#3     3       4       3     3
#4     4       2       2     1

For 2nd step to calculate the percentage you can do:

result %>%
  summarise(across(Tobacco:PA, ~mean(. > 1) * 100))

#  Tobacco Alcohol    PA
#    <dbl>   <dbl> <dbl>
#1    0.75    0.75   0.5

Answer 2

In base R , we can do

aggregate(.~ pt_id, mydata, FUN = function(x) sum(!is.na(x)), na.action = NULL)

-output

#   pt_id Tobacco Alcohol PA
#1     1       3       3  2
#2     2       0       1  0
#3     3       4       3  3
#4     4       2       2  1

---

Or more compactly with rowsum from base R

rowsum(+(!is.na(mydata[-1])), mydata$pt_id)
#  Tobacco Alcohol PA
#1       3       3  2
#2       0       1  0
#3       4       3  3
#4       2       2  1

---

If we need the percentages

colMeans(rowsum(+(!is.na(mydata[-1])), mydata$pt_id) > 1)
#Tobacco Alcohol      PA 
#   0.75    0.75    0.50