CODEWITHSUNDEEP
rdplyrpivotdata-manipulation
principe
principe

Reputation: 303

Pivot longer multiple columns while pivot wider others

Hello I have a dataset of 3-5 rows per group as the following and I want to put some columns in a longer format and have on column in a wider format.

The first dataset below represents the original format and I want to transform it to the second one. I have used pivot wider cols = c("Jan", "Feb") but i can't manage to pivot the Type column to be longer at the same time.

data <- as.data.frame(matrix(ncol=5, nrow=6))
colnames(data) <- c("names", "group", "Type", "Jan", "Feb")
data$names <- c("P1", "P1", "P1", "P2", "P2", "P2")
data$group <- "S"
data$Type <- c("Beg", "Middle", "End", "Beg", "Middle", "End")
data$Jan <- c(1, 2, 3, 10, 5, 15)
data$Feb <- c(5, 5, 10, 5, 2, 7)

    
   names group Type     Jan  Feb
1   P1    S    Beg       1   5
2   P1    S    Middle    2   5
3   P1    S    End       3   10
4   P2    S    Beg       10  5
5   P2    S    Middle    5   2
6   P2    S    End       15  7


data_transformed <- as.data.frame(matrix(ncol=6, nrow=4))
colnames(data_transformed) <- c("names", "group", "Month", "Beg", "Middle", "End")
data_transformed$names <- c("P1", "P1", "P2", "P2")
data_transformed$group <- "S"
data_transformed$Month <- c("Jan", "Feb")
data_transformed$Beg <- c(1, 10, 5, 5)
data_transformed$Middle <- c(2, 5, 5, 2)
data_transformed$End <- c(2, 15, 10, 7)

  names group Month   Beg Middle End
1   P1  S     Jan      1    2    2
2   P1  S     Feb      10   5    15
3   P2  S     Jan      5    5    10
4   P2  S     Feb      5    2    7

Upvotes: 3

Views: 1260

Answers (3)

G. Grothendieck
G. Grothendieck

Reputation: 270045

This is several years too late but proc_transpose in the procs package (which may not have existed on CRAN at the time) can transpose by group.

Below the code specifies the columns to group by (by=). The id column (id=) is the input column that becomes the column names in the output data frame. Here it need not be specified since there is only one character column left after removing the two grouping columns and the default is to assume that a single character column is the id column. The column name of the new column (name=) in the output that holds the column names from the input is specified as "Month" but could be omitted if the default of NAME were adequate.

library(procs)

proc_transpose(data, by = c("names", "group"), name = "Month")
##   names group Month Beg Middle End
## 1    P1     S   Jan   1      2   3
## 2    P1     S   Feb   5      5  10
## 3    P2     S   Jan  10      5  15
## 4    P2     S   Feb   5      2   7

Upvotes: 1

ThomasIsCoding
ThomasIsCoding

Reputation: 102529

A data.table option using dcast + melt

dcast(
  melt(
    setDT(data),
    id.vars = c("names", "group", "Type"),
    variable.name = "Month"
  ),
  names + group + Month ~ Type
)

gives

   names group Month Beg End Middle
1:    P1     S   Jan   1   3      2
2:    P1     S   Feb   5  10      5
3:    P2     S   Jan  10  15      5
4:    P2     S   Feb   5   7      2

Upvotes: 1

akrun
akrun

Reputation: 887711

Here, we need a pivot_longer + pivot_wider i.e. first reshape to 'long' with cols Jan to 'Feb', then reshape the long back to wider format with column names from 'Type'

library(dplyr)
library(tidyr)
data %>%
     pivot_longer(cols = Jan:Feb, names_to = 'Month') %>% 
     pivot_wider(names_from = Type, values_from = value)

-output

# A tibble: 4 x 6
#  names group Month   Beg Middle   End
#  <chr> <chr> <chr> <dbl>  <dbl> <dbl>
#1 P1    S     Jan       1      2     3
#2 P1    S     Feb       5      5    10
#3 P2    S     Jan      10      5    15
#4 P2    S     Feb       5      2     7

Or use recast from reshape2

library(reshape2)
recast(data, measure = c("Jan", "Feb"),
     names + group + variable ~ Type, values.var = 'value')

Upvotes: 3

Related Questions