Collapsing Dataframe Rows along several variables - r

Question

I have a dataframe that looks something like this, in which I have several rows for each user, and many NAs in the columns.
user
Effect T1
Effect T2
Effect T3
Benchmark T1
Benchmark T2
Benchmark T3
Tom
01
NA
NA
02
NA
NA
Tom
NA
07
NA
NA
08
NA
Tom
NA
NA
13
NA
NA
14
Larry
03
NA
NA
04
NA
NA
Larry
NA
09
NA
NA
10
NA
Larry
NA
NA
15
NA
NA
16
Dave
05
NA
NA
06
NA
NA
Dave
NA
11
NA
NA
12
NA
Dave
NA
NA
17
NA
NA
18
I want to collapse the columns using the name and filling the values from reach row, this this.
user
Effect T1
Effect T2
Effect T3
Benchmark T1
Benchmark T2
Benchmark T3
Tom
01
07
13
02
08
14
Larry
03
09
15
04
10
16
Dave
05
11
17
06
12
18
How might I accomplish this?
Thank you in advance for your help. Update: I've added the dput of a subset of the actual data below.
structure(list(name = c("Abraham_Ralph", "Abraham_Ralph", "Abraham_Ralph",
"Ackerman_Gary", "Adams_Alma", "Adams_Alma", "Adams_Alma", "Adams_Alma",
"Adams_Sandy", "Aderholt_Robert", "Aderholt_Robert", "Aderholt_Robert",
"Aderholt_Robert", "Aderholt_Robert", "Aguilar_Pete", "Aguilar_Pete",
"Aguilar_Pete"), state = c("LA", "LA", "LA", "NY", "NC", "NC",
"NC", "NC", "FL", "AL", "AL", "AL", "AL", "AL", "CA", "CA", "CA"
), seniority = c(1, 2, 3, 15, 1, 2, 3, 4, 1, 8, 9, 10, 11, 12,
1, 2, 3), legeffect_112 = c(NA, NA, NA, 0.202061712741852, NA,
NA, NA, NA, 1.30758035182953, 3.73544979095459, NA, NA, NA, NA,
NA, NA, NA), legeffect_113 = c(NA, NA, NA, NA, 0, NA, NA, NA,
NA, NA, 0.908495426177979, NA, NA, NA, NA, NA, NA), legeffect_114 = c(2.07501077651978,
NA, NA, NA, NA, 0.84164834022522, NA, NA, NA, NA, NA, 0.340001106262207,
NA, NA, 0.10985741019249, NA, NA), legeffect_115 = c(NA, 0.493490308523178,
NA, NA, NA, NA, 0.587624311447144, NA, NA, NA, NA, NA, 0.159877583384514,
NA, NA, 0.730929613113403, NA), legeffect_116 = c(NA, NA, 0.0397605448961258,
NA, NA, NA, NA, 1.78378939628601, NA, NA, NA, NA, NA, 0.0198802724480629,
NA, NA, 0.0497006773948669), benchmark_112 = c(NA, NA, NA, 0.738679468631744,
NA, NA, NA, NA, 0.82908970117569, 1.39835929870605, NA, NA, NA,
NA, NA, NA, NA), benchmark_113 = c(NA, NA, NA, NA, 0.391001850366592,
NA, NA, NA, NA, NA, 1.58223271369934, NA, NA, NA, NA, NA, NA),
benchmark_114 = c(1.40446054935455, NA, NA, NA, NA, 0.576326191425323,
NA, NA, NA, NA, NA, 1.42212760448456, NA, NA, 0.574363172054291,
NA, NA), benchmark_115 = c(NA, 1.3291300535202, NA, NA, NA,
NA, 0.537361204624176, NA, NA, NA, NA, NA, 1.45703768730164,
NA, NA, 0.523149251937866, NA), benchmark_116 = c(NA, NA,
0.483340591192245, NA, NA, NA, NA, 1.31058621406555, NA,
NA, NA, NA, NA, 0.751261711120605, NA, NA, 1.05683290958405
)), row.names = c(NA, -17L), class = c("tbl_df", "tbl", "data.frame"
))

Answer 1

A data.table solution:
# melt data, remove NA, then recast ...
dt <- dcast(melt(data.table(d), "name")[!value %in% NA], name ~ variable)
dcast(melt(data.table(d), "name")[!value %in% c(NA) & !variable %in% c("variable", "seniority", "state")], name ~ variable)
name legeffect_112 legeffect_113 legeffect_114 legeffect_115 legeffect_116 benchmark_112 benchmark_113 benchmark_114 benchmark_115 benchmark_116
1: Abraham_Ralph <NA> <NA> 2.07501077651978 0.493490308523178 0.0397605448961258 <NA> <NA> 1.40446054935455 1.3291300535202 0.483340591192245
2: Ackerman_Gary 0.202061712741852 <NA> <NA> <NA> <NA> 0.738679468631744 <NA> <NA> <NA> <NA>
3: Adams_Alma <NA> 0 0.84164834022522 0.587624311447144 1.78378939628601 <NA> 0.391001850366592 0.576326191425323 0.537361204624176 1.31058621406555
4: Adams_Sandy 1.30758035182953 <NA> <NA> <NA> <NA> 0.82908970117569 <NA> <NA> <NA> <NA>
5: Aderholt_Robert 3.73544979095459 0.908495426177979 0.340001106262207 0.159877583384514 0.0198802724480629 1.39835929870605 1.58223271369934 1.42212760448456 1.45703768730164 0.751261711120605
6: Aguilar_Pete <NA> <NA> 0.10985741019249 0.730929613113403 0.0497006773948669 <NA> <NA> 0.574363172054291 0.523149251937866 1.05683290958405
Data/Setup
# Load data.table
# install.packages("data.table")
library(data.table)
# Read example data
d <- structure(list(name = c("Abraham_Ralph", "Abraham_Ralph", "Abraham_Ralph",
"Ackerman_Gary", "Adams_Alma", "Adams_Alma", "Adams_Alma", "Adams_Alma",
"Adams_Sandy", "Aderholt_Robert", "Aderholt_Robert", "Aderholt_Robert",
"Aderholt_Robert", "Aderholt_Robert", "Aguilar_Pete", "Aguilar_Pete",
"Aguilar_Pete"), state = c("LA", "LA", "LA", "NY", "NC", "NC",
"NC", "NC", "FL", "AL", "AL", "AL", "AL", "AL", "CA", "CA", "CA"
), seniority = c(1, 2, 3, 15, 1, 2, 3, 4, 1, 8, 9, 10, 11, 12,
1, 2, 3), legeffect_112 = c(NA, NA, NA, 0.202061712741852, NA,
NA, NA, NA, 1.30758035182953, 3.73544979095459, NA, NA, NA, NA,
NA, NA, NA), legeffect_113 = c(NA, NA, NA, NA, 0, NA, NA, NA,
NA, NA, 0.908495426177979, NA, NA, NA, NA, NA, NA), legeffect_114 = c(2.07501077651978,
NA, NA, NA, NA, 0.84164834022522, NA, NA, NA, NA, NA, 0.340001106262207,
NA, NA, 0.10985741019249, NA, NA), legeffect_115 = c(NA, 0.493490308523178,
NA, NA, NA, NA, 0.587624311447144, NA, NA, NA, NA, NA, 0.159877583384514,
NA, NA, 0.730929613113403, NA), legeffect_116 = c(NA, NA, 0.0397605448961258,
NA, NA, NA, NA, 1.78378939628601, NA, NA, NA, NA, NA, 0.0198802724480629,
NA, NA, 0.0497006773948669), benchmark_112 = c(NA, NA, NA, 0.738679468631744,
NA, NA, NA, NA, 0.82908970117569, 1.39835929870605, NA, NA, NA,
NA, NA, NA, NA), benchmark_113 = c(NA, NA, NA, NA, 0.391001850366592,
NA, NA, NA, NA, NA, 1.58223271369934, NA, NA, NA, NA, NA, NA),
benchmark_114 = c(1.40446054935455, NA, NA, NA, NA, 0.576326191425323,
NA, NA, NA, NA, NA, 1.42212760448456, NA, NA, 0.574363172054291,
NA, NA), benchmark_115 = c(NA, 1.3291300535202, NA, NA, NA,
NA, 0.537361204624176, NA, NA, NA, NA, NA, 1.45703768730164,
NA, NA, 0.523149251937866, NA), benchmark_116 = c(NA, NA,
0.483340591192245, NA, NA, NA, NA, 1.31058621406555, NA,
NA, NA, NA, NA, 0.751261711120605, NA, NA, 1.05683290958405
)), row.names = c(NA, -17L), class = c("tbl_df", "tbl", "data.frame"
))

Answer 2

This solution is using only the base functions (no extra packages), but the one-liner may cause eyes to cross, so I'll split it into several functions.
The plan is the following:
Split the original data.frame by the values in name column, using the function by;
For each partition of the data.frame, collapse the columns;
A collapsed column returns the max value of the column, or NA if all its values are NA;
The collapsed data.frame partitions are stacked together.
So, this is a function that does that:
dfr_collapse <- function(dfr, col0)
{
# Collapse the columns of the data.frame "dfr" grouped by the values of
# the column "col0"
# Max/NA function
namax <- function(x)
{
if(all(is.na(x)))
NA # !!!
else
max(x, na.rm=TRUE)
}
# Column collapse function
byfun <- function(x)
{
lapply(x, namax)
}
# Stack the partitioning results
return(do.call(
what = rbind,
args = by(dfr, dfr[[col0]], byfun)
))
}
May not look as slick as a one-liner, but it does the job. It can be tunrned into a one-liner, but you don't want that.
Assuming that df0 is the data.frame from you dput, you can test this function with
dfr_collapse(df0)
Nota bene: for the sake of simplicity, I return an NA of type logical (see the comment # !!! above). The correct code should convert that NA to the mode of the x vector. Also, the function should check the type of its inputs, etc.