r - deciding on the number of significant digits in a data frame

Question

I have a huge data frame, a sample of 3 columns and 11 rows is given below:

df <- structure(list(A = c(61960, 273, 439, 38877, 75325, 80929, 
23028, 57240, 10140, 25775, 7286), B = c(10, 12, 11, 13, 2, 1, 1, 
1, 1, 1, 1), C = c(122, 140, 163, 12, 190, 16, 14, 18, 15, 17, 16
)), class = c("tbl_df", "tbl", "data.frame"), row.names = c(NA, 
-11L))

For each column of the data frame, I would like to calculate the median number of significant digits for each order of magnitude in that column.

So for example, for column A above, there are 3 orders of magnitude present (10^3, 10^4, 10^5). The first number has 4 digits (last zero doesn't count), second has 3, and so on.

My output should be a list for each column, with one element a vector containing the orders of magnitude, and the second the median number of significant digits. So for each column I am expecting a list, my output would be a list of lists. For example for column A:

L[["A"]] = list(c(5,4,3), c(5, 4, 3))

Why is this the list? In column A there are 3 different orders of magnitude: 10^5, 10^4, 10^3. The median number of significant digits for the 10^5 o.o.m is 5, for 10^4, 4, and for 10^3, 3.

Is there a way to do this efficiently? with something like mutate or map (not apply, because this would be the same as using a loop).

标签： rdataframevectorization

Answer 1

We can do this by looping over the columns, then grouped by the nchar of the column, remove the 0s at the end with sub, get the median and return a list of the median along with the grouping variable in tapply (returned as the names of the named vector)

lapply(df, function(x) {
      x1 <- tapply(nchar(sub("0+$", "", x)), nchar(x), FUN = median )
      list(as.integer(names(x1)), as.numeric(x1))

   })
#$A
#$A[[1]]
#[1] 3 4 5

#$A[[2]]
#[1] 3 4 5


#$B
#$B[[1]]
#[1] 1 2

#$B[[2]]
#[1] 1 2


#$C
#$C[[1]]
#[1] 2 3

#$C[[2]]
#[1] 2.0 2.5

---

Or this can be also done with tidyverse and return as a single dataset

library(tidyverse)
df %>%
   mutate_all(str_remove, "0+$") %>%
   map2_dfr(., df,  ~ 
        tibble(x = nchar(.x), grp = nchar(.y)) %>% 
          group_by(grp) %>%
          summarise(x = median(x)), .id = 'colName')
# A tibble: 7 x 3
#  colName   grp     x
#  <chr>   <int> <dbl>
#1 A           3   3  
#2 A           4   4  
#3 A           5   5  
#4 B           1   1  
#5 B           2   2  
#6 C           2   2  
#7 C           3   2.5