CODEWITHSUNDEEP
rdictionarylapplyradix
sippstress
sippstress

Reputation: 33

Using column value as dataframe index in list dataframes (Map or lapply with seq_along)?

I have a list of dataframes list1 and need a new column 'mn' in each dataframe that is the mean of a conditional number of columns based on the value in another column num plus one. So, for num=3 the new column would be the mean of the first four columns. For the example below

df1 <- data.frame(num= c(3, 1, 1, 1, 2), d1= c(1, 17, 17, 17, 15), d2= c(1, 15, 15, 15, 21), d3= c(6, 21, 21, 21, 23), d4= c(2, 3, 3, 3, 2))
df2 <- data.frame(num= c(3, 2, 2, 2, 2), d1= c(1, 10, 10, 10, 15), d2= c(1, 5, 5, 5, 21), d3= c(6, 2, 2, 2, 23), d4= c(2, 3, 3, 3, 5))
list1 <- list(df1, df2)

I would expect

newlist
[[1]]
   num   d1   d2   d3   d4    mn
1   3    1    1    6    2    2.5
2   1    17   15   21   3    16.0  
3   1    17   15   21   3    16.0

The closest I've gotten is

newlist <- lapply(list1, function(x) {
  x <- cbind(x, sapply(x$num, function(y) {
      y <- rowSums(x[2:(2+y)])/(y+1)
      }))
  })

which binds columns for the means of every row. Based on this post I think I need a seq_along or maybe a Map on the inside function but I can't figure out how to implement it.

Upvotes: 1

Views: 94

Answers (1)

akrun
akrun

Reputation: 887511

An option is to loop over the list with lapply, extract the number of elements for each row with apply based on the 'num' column value (+1), get the mean and create the new column in transform

lapply(list1, function(x)  transform(x,
       mn = apply(x, 1, function(y) mean(y[-1][seq(y[1]+1)]))))
#[[1]]
#  num d1 d2 d3 d4       mn
#1   3  1  1  6  2  2.50000
#2   1 17 15 21  3 16.00000
#3   1 17 15 21  3 16.00000
#4   1 17 15 21  3 16.00000
#5   2 15 21 23  2 19.66667

#[[2]]
#  num d1 d2 d3 d4        mn
#1   3  1  1  6  2  2.500000
#2   2 10  5  2  3  5.666667
#3   2 10  5  2  3  5.666667
#4   2 10  5  2  3  5.666667
#5   2 15 21 23  5 19.666667

Or with tidyverse, by pivoting to 'long' format with pivot_longer, do a group by row and get the mean of the first 'n' elements based on the 'num' value

library(purrr)
library(dplyr)
library(tidyr)
map(list1, ~
        .x %>% 
           mutate(rn = row_number()) %>%
           pivot_longer(cols = starts_with('d')) %>% 
           group_by(rn) %>% 
           summarise(value = mean(value[seq_len(first(num) + 1)])) %>%
           pull(value) %>%
           bind_cols(.x, mn = .))

Upvotes: 1

Related Questions