How to drop NA variables in a data frame by row - r

Question

Here is my data frame:
structure(list(Q = c(NA, 346.86, 166.95, 162.57, NA, NA, NA,
266.7), L = c(18.93, NA, 15.72, 39.51, NA, NA, NA, NA), C = c(NA,
23.8, NA, 8.47, 20.89, 18.72, 14.94, NA), X = c(40.56, NA, 26.05,
3.08, 23.77, 59.37, NA, NA), W = c(29.47, NA, NA, NA, 36.08,
NA, 27.34, 28.19), S = c(NA, 7.47, NA, NA, 18.64, NA, 25.34,
NA), Y = c(NA, 2.81, 0, NA, NA, 21.18, 10.83, 12.19), H = c(0,
NA, NA, NA, NA, 0, NA, 0)), class = "data.frame", row.names = c(NA,
-8L), .Names = c("Q", "L", "C", "X", "W", "S", "Y", "H"))
Each row has 4 variables that are NAs, now I want to do the same operations to every row:
Drop those 4 varibles that are NAs
Calculate diversity for the rest 4 variables (it's just some computations involved with the rest, here I use diversity() from vegan)
Append the output to a new data frame
But the problem is:
How to do drop NA variables using dplyr? I don't know whether select() can make it.
How to apply operations to every row of a data frame?
It seems that drop_na() will remove the entire row for my dataset, any suggestion?

Answer 1

With tidyverse it may be better to gather into 'long' format and then spread it back. Assuming that we have exactly 4 non-NA elements per row, create a row index with rownames_to_column (from tibble), gather (from tidyr) into 'long' format, remove the NA elements, grouped by row number ('rn'), change the 'key' values to common values and then spread it to wide' format
library(tibble)
library(tidyr)
library(dplyr)
res <- rownames_to_column(df1, 'rn') %>%
gather(key, val, -rn) %>%
filter(!is.na(val)) %>%
group_by(rn) %>%
mutate(key = LETTERS[1:4]) %>%
spread(key, val) %>%
ungroup %>%
select(-rn)
res
# A tibble: 8 x 4
# A B C D
#* <dbl> <dbl> <dbl> <dbl>
#1 18.9 40.6 29.5 0
#2 347 23.8 7.47 2.81
#3 167 15.7 26.0 0
#4 163 39.5 8.47 3.08
#5 20.9 23.8 36.1 18.6
#6 18.7 59.4 21.2 0
#7 14.9 27.3 25.3 10.8
#8 267 28.2 12.2 0
diversity(res)
# 1 2 3 4 5 6 7 8
#1.0533711 0.3718959 0.6331070 0.7090783 1.3517680 0.9516232 1.3215712 0.4697572
Regarding the diversity calculation, we can replace NA with 0 and apply on the whole dataset i.e.
library(vegan)
diversity(replace(df1, is.na(df1), 0))
#[1] 1.0533711 0.3718959 0.6331070 0.7090783
#[5] 1.3517680 0.9516232 1.3215712 0.4697572
as we get the same output as in the first solution