conditionally transpose select rows using tidyverse

Question

I have a dataset that I'm working with that I'm attempting to reshape using tidyverse.

From:

|Name |eval   |test |type      | score|
|:----|:------|:----|:---------|-----:|
|John |first  |1    |pretest   |    10|
|John |first  |1    |posttest  |    15|
|John |first  |2    |pretest   |    20|
|John |first  |2    |posttest |    30|
|John |second |1    |pretest   |    35|
|John |second |1    |posttest  |    50|
|John |second |2    |pretest   |     5|
|John |second |2    |posttest |    10|
|Jane |first  |1    |pretest   |    40|
|Jane |first  |1    |posttest  |    20|
|Jane |first  |2    |pretest   |    10|
|Jane |first  |2    |posttest |    20|

To:

|Name |eval   |new_name      | pre_test| post_test|
|:----|:------|:-------------|--------:|---------:|
|John |first  |John_first_1  |       10|        15|
|John |first  |John_first_2  |       20|        30|
|John |second |John_second_1 |       35|        50|
|John |second |John_second_2 |        5|        10|
|Jane |first  |Jane_first_1  |       40|        20|
|Jane |first  |Jane_first_2  |       10|        20|

• tried doing group_by in order to group_by Name, eval, and test so that each group would essentially be pre_test vs. post_test for a given person.

• also tried using unite on Name, eval, test, and type. But if I do a spread after that then each the unique name end up being a number of columns.

• also tried to doing a unite first on Name, eval, test first, and then a spread using key=(new united name) and value =Value, but the output isn't what I wanted

I know a loop function can be written to take every other value and put into a new column, but I'm trying to see if there's a tidyverse way to go about this.

Thanks!!

library(tidyverse)
Name <- c('John', 'John', 'John', 'John',
              'John', 'John', 'John', 'John',
              'Jane', 'Jane', 'Jane', 'Jane')
eval <- c('first', 'first', 'first', 'first',
          'second', 'second', 'second', 'second',
          'first', 'first', 'first', 'first')
test <- c('1', '1', '2', '2',
          '1', '1', '2', '2',
          '1', '1', '2', '2')
type <- c('pretest', 'posttest', 'pretest', 'posttest',
          'pretest', 'posttest', 'pretest', 'posttest',
          'pretest', 'posttest', 'pretest', 'posttest')
score <- c(10, 15, 20, 30, 35, 50, 5, 10, 40, 20, 10, 20)
df <- data.frame(Name, eval, test, type, score)

df %>%
  unite(temp, Name, eval, test) %>%
  spread(key=type, value=score)

Edit to show the original table that akrun's code worked on From:

|Name |eval   |test |type      | score|
|:----|:------|:----|:---------|-----:|
|John |first  |1    |pretest   |    10|
|John |first  |1    |posttest  |    15|
|John |first  |2    |pretest   |    20|
|John |first  |2    |postttest |    30|
|John |second |1    |pretest   |    35|
|John |second |1    |posttest  |    50|
|John |second |2    |pretest   |     5|
|John |second |2    |postttest |    10|
|Jane |first  |1    |pretest   |    40|
|Jane |first  |1    |posttest  |    20|
|Jane |first  |2    |pretest   |    10|
|Jane |first  |2    |postttest |    20|

Answer 1

We can replace the multiple 't's in the 'type' column to make it same, then use unite specify the remove = FALSE to keep the initial columns as well and spread

library(dplyr)
library(tidyr)
library(stringr)
df %>% 
   mutate(type = str_replace(type, "t{2,}", "t")) %>%
   unite(new_name, Name, eval, test, remove = FALSE) %>% 
   spread(type, score)
#       new_name Name   eval test postest pretest
#1  Jane_first_1 Jane  first    1      20      40
#2  Jane_first_2 Jane  first    2      20      10
#3  John_first_1 John  first    1      15      10
#4  John_first_2 John  first    2      30      20
#5 John_second_1 John second    1      50      35
#6 John_second_2 John second    2      10       5

In the new version tidyr_1.0.0, pivot_wider is introduced and it can be used as a more generalized version of spread (would be deprecated in the future). So, instead of the spread line at the end, use

 ...%>%
    pivot_wider(names_from = type, values_from = score)

Answer 2

How about something like....

data <- tibble(
  Name = c(rep("John", 8), rep("Jane", 4)),
  eval = c(rep("first", 4), rep("second", 4), rep("first", 4)),
  type = rep(c("pretest", "posttest"), 6),
  score = c(10, 15, 20, 30, 35, 50, 5, 10, 40, 20, 10, 20)
)

data %>% 
  group_by(Name, eval, type) %>% 
  mutate(num = 1:n(),
         new_name = str_c(Name, "_", eval, "_", num)) %>% 
  ungroup()  %>% 
  dplyr::select(new_name, type, score) %>% 
  spread(type, score) 

Which yields:

# A tibble: 6 x 3
  new_name      posttest pretest
  <chr>            <dbl>   <dbl>
1 Jane_first_1        20      40
2 Jane_first_2        20      10
3 John_first_1        15      10
4 John_first_2        30      20
5 John_second_1       50      35
6 John_second_2       10       5