r tidyverse data-conversion multilevel-analysis

How to convert a wide data into long format for cross-classified model [R, GLMM]

I would like to convert wide data to long data in R, and my data set is for cross-classified models, exploring participants’ response to each target item that has different characteristics.

condition is one of the two conditions where participants were assigned to.
The participants were tested twice: t1 and t2.
As for item-level predictor variables, x1 and x2, are coded.
As for response, whether participants’ response to the item was right or wrong was coded.
two test formats were administered, test1 and test2.

Although there are so many tutorials for a wide to long conversion, I could not find a one specifically explaining conversion for cross-classified models.

I would like to use tidyverse if possible for the sake of consistency.

My sample data is the following:

structure(list(item_name = c("x1", "x2", "participant_id", "1", 
"2", "3", "4", "5", "6", "7"), participant_variable_1 = c(NA, 
NA, NA, 20, 23, 21, 20, 19, 22, 30), condition = c(NA, NA, NA, 
"A", "B", "A", "B", "A", "B", "A"), t1.item1.test1 = c(1, 3, 
NA, 0, 1, 0, 1, 0, 0, 1), t1.item2.test1 = c(2, 2, NA, 0, 0, 
0, 1, 1, 0, 1), t1.item3.test1 = c(1, 3, NA, 0, 0, 0, 1, 0, 0, 
0), t1.item4.test1 = c(3, 1, NA, 1, 0, 0, 0, 1, 1, 0), t2.item1.test1 = c(1, 
3, NA, 0, 1, 1, 0, 1, 1, 1), t2.item2.test1 = c(2, 2, NA, 1, 
0, 1, 0, 1, 0, 1), t2.item3.test1 = c(1, 3, NA, 0, 0, 0, 1, 0, 
0, 0), t2.item4.test1 = c(3, 1, NA, 1, 1, 0, 1, 1, 1, 0), t1.item1.test2 = c(1, 
3, NA, 0, 1, 0, 1, 0, 0, 1), t1.item2.test2 = c(2, 2, NA, 0, 
0, 0, 1, 1, 0, 1), t1.item3.test2 = c(1, 3, NA, 0, 0, 0, 1, 0, 
0, 0), t1.item4.test2 = c(3, 1, NA, 1, 0, 0, 0, 1, 1, 0), t2.item1.test2 = c(1, 
3, NA, 0, 1, 1, 0, 1, 1, 1), t2.item2.test2 = c(2, 2, NA, 1, 
0, 1, 0, 1, 0, 1), t2.item3.test2 = c(1, 3, NA, 0, 0, 0, 1, 0, 
0, 0), t2.item4.test2 = c(3, 1, NA, 1, 1, 0, 1, 1, 1, 0)), row.names = c(NA, 
-10L), class = c("tbl_df", "tbl", "data.frame"))

I would like to have a long data, which looks like the following:

Please and thank you for your guidance!

Solution

This answer requires heavy use of the new pivot_ functions in the dev version of tidyr. You can install that with devtools::install_github("tidyverse/tidyr") if you're willing to run the dev version.

First we split the data into item and participant info - you're not really getting any benefit from storing both in the same table:

item_info = dat[1:2, ]
participant_info = dat[4:nrow(dat), ] %>%
    rename(participant_id = item_name)

Then it's time for a lot of pivoting:

# I have the dev version of tidyr so that is being loaded
library(tidyverse)

item_long = item_info %>%
    select(-participant_variable_1, -condition) %>%
    pivot_longer(
        cols = t1.item1:t2.item4,
        names_to = c("time", "item"),
        names_pattern = "t(\\d)\\.(item\\d)",
    ) %>%
    pivot_wider(names_from = item_name, values_from = value)

participant_long = participant_info %>%
    pivot_longer(
        cols = t1.item1:t2.item4,
        names_to = c("time", "item"),
        names_pattern = "t(\\d)\\.(item\\d)",
        values_to = "response"
    )

combined = participant_long %>%
    left_join(item_long, by = c("item", "time"))

Result:

> combined
# A tibble: 56 x 8
   participant_id participant_variable_1 condition time  item  response    x1    x2
   <chr>                           <dbl> <chr>     <chr> <chr>    <dbl> <dbl> <dbl>
 1 1                                  20 A         1     item1        0     1     3
 2 1                                  20 A         1     item2        0     2     2
 3 1                                  20 A         1     item3        0     1     3
 4 1                                  20 A         1     item4        1     3     1