Separate columns with constant numbers and condens

Question

I have a data.frame called d. In this data.frame, some columns consist of constant numbers across the rows of the first column: study.name (see below).

For example, columns ESL, ESL.1, prof, and prof.1 are constant numbers for all rows of Shin.Ellis and also constant for all rows of Trus.Hsu and so on.

Q: In BASE R, how can I separate such constant variables, and then condense them to one row with only one number?

My desired output is shown further below. A functional answer is appreciated.

d <- read.csv("https://raw.githubusercontent.com/izeh/m/master/irr.csv", h = T)[-(2:3)]

## FIRST 8 ROWS:

#    study.name ESL prof scope type ESL.1 prof.1 scope.1 type.1
# 1  Shin.Ellis   1    2     1    1     1      2       1      1
# 2  Shin.Ellis   1    2     1    1     1      2       1      1
# 3  Shin.Ellis   1    2     1    2     1      2       1      1
# 4  Shin.Ellis   1    2     1    2     1      2       1      1
# 5  Shin.Ellis   1    2    NA   NA     1      2      NA     NA
# 6  Shin.Ellis   1    2    NA   NA     1      2      NA     NA
# 7    Trus.Hsu   2    2     2    1     2      2       1      1
# 8    Trus.Hsu   2    2    NA   NA     2      2      NA     NA

Desired output:

#    study.name ESL prof  ESL.1 prof.1 
# 1  Shin.Ellis   1    2      1      2  
# 2  Trus.Hsu     2    2      2      2
# .     .         .    .      .      . # AND SO ON !!!

Answer 1

You could try something like this, though it feels a bit clumsy. Basically, check which columns have constant values by group for all groups, keep only those columns, and then keep only unique values (since now they are constant by group).

d <- read.table(header = TRUE,
text = "study.name ESL prof scope type ESL.1 prof.1 scope.1 type.1
Shin.Ellis   1    2     1    1     1      2       1      1
Shin.Ellis   1    2     1    1     1      2       1      1
Shin.Ellis   1    2     1    2     1      2       1      1
Shin.Ellis   1    2     1    2     1      2       1      1
Shin.Ellis   1    2    NA   NA     1      2      NA     NA
Shin.Ellis   1    2    NA   NA     1      2      NA     NA
  Trus.Hsu   2    2     2    1     2      2       1      1
  Trus.Hsu   2    2    NA   NA     2      2      NA     NA")

is_constant <- function(x) length(unique(x)) == 1L

keep_constants <- function(df, group_col) {
  data_cols <- colnames(df)[setdiff(1:ncol(df), group_col)]
  check_df <- aggregate(df, by = list(df[[group_col]]), FUN = is_constant)
  cols_to_keep <- sapply(check_df[, -1], all)
  unique(df[, cols_to_keep])
}

keep_constants(d, 1)
#>   study.name ESL prof ESL.1 prof.1
#> 1 Shin.Ellis   1    2     1      2
#> 7   Trus.Hsu   2    2     2      2

^{Created on 2019-10-09 by the reprex package (v0.3.0)}

Answer 2

If you just want to remove repeated values across all columns unique() is base R

unique(d)

EDIT - Thanks for the clarification @CalumYou - I think this is what OP is looking for in base R.

is_constant = lapply(split(d, d$study.name), function(data){
  unlist(lapply(data,function(col){
    length(unique(col)) == 1
  }))
})
is_constant = as.data.frame(do.call(rbind, is_constant))
all_constant = d[,unlist(lapply(is_constant,all))]
all_constant = unique(all_constant)

Answer 3

May be we need

library(dplyr)
d %>%
   group_by(study.name) %>%
   slice(1)

---

Or in base R after grouping by 'study.name', get the first row while specifying the na.action = NULL as the default option is na.omit which can omit any row having NA in any of the columns

aggregate(.~ study.name, d, head, 1, na.action = NULL)

If we want to subset the columns

nm1 <- names(which(!colSums(!do.call(rbind, by(d[-1], d$study.name,
     FUN = function(x) lengths(sapply(x, unique)) == 1)))))
unique(d[c("study.name", nm1)])

Answer 4

d_list <- lapply(split(d,d$study.name), 
                 #Find columns with similar values using sapply and length(unique(cols)) 
                 #then get the 1st row
                 function(x) x[1, sapply(x,function(y) length(unique(y))==1)])
do.call('rbind.data.frame',d_list)

           study.name ESL prof ESL.1 prof.1
Bit.KnoA     Bit.KnoA   1    3     1      3
Bit.KnoB     Bit.KnoB   1    2     1      2
ChandlerA   ChandlerA   1    2     1      2
Mubarak       Mubarak   2   NA     2     NA
SheenA         SheenA   1    2     1      2
Shin.Ellis Shin.Ellis   1    2     1      2
Sun               Sun   2    2     2      2
Trus.Hsu     Trus.Hsu   2    2     2      2