Converting wide to long data in R

1. Create a subject id

To make this transformation we will use the tidyverse functions. First, we will want to keep track of our 10 subjects so we will create a column “id” with an integer to identify them, then bind that column to the wide data to create a new dataframe.

#1. Create a new data frame with a subject id
long.data <- cbind(id = c(1:10), chapter_12_table_1)

2. Gather dependent variables

Next, we will use the gather() function which will stack each column of the wide data under a column called reaction_time and create a new column that will add the column name of the wide data as a value in a new column called condition.angle. The -id is used to prevent getting the id number into the condition.angle column.

long.data %>% gather(key = condition.angle,
                     value = reaction_time, -id)

3. Separate the condition.angle column

So far we’re on a good track, but the column condition.angle needs to be separated into two columns. One to specify condition, the other to specify angle. The separate() function can help us out here by splitting the condition.angle column into two columns, condition and angle. The sep = -1 argument tells the separate() function to split the values 1 character from the end. This separates the last character which contains a single digit angle number into a separate column. Try substituting different numbers such as -2 or 1, to get a feel for what the function is doing here.

long.data %>% 
  gather(key = condition.angle,
                     value = reaction_time, 
                     -id) %>%
  separate(col = condition.angle, 
           into = c("condition", "angle"), 
           sep = -1)

4. Sort the data

At this point, we’ve done most of what is needed to convert the wide data to long data. The next step in the series of tidyverse functions is to sort the data. To get the output displayed above, we will want to sort the data by subject id, then by condition, then by angle. To accomplish this we will use the arrange() function and will display the reaction_time values as a column of the transposed rows from the wide data set. We will re-save this manipulation as long.data. Notice how the number of rows is now 60 whereas we started with 10.

long.data <- long.data %>% 
  gather(key = condition.angle,
         value = reaction_time, 
         -id) %>%
  separate(col = condition.angle, 
           into = c("condition", "angle"), 
           sep = -1) %>%
  arrange(id, 
          condition, 
          angle)

5. Convert to factor

Now before conducting any statistical tests, we will want to convert our id, condition, and angle columns to factor. We can do this with the base R function as.factor().

long.data$id <- as.factor(long.data$id)
long.data$condition <- as.factor(long.data$condition)
long.data$angle <- as.factor(long.data$angle)

Bonus with other helper functions

We could also convert variable to factor with the convert_as_factor() function from the rstatix package. We can pipe each of the steps above and add the convert_as_factor() function at the end setting the columns we would like to convert.

library(rstatix)
# Create a new data frame with a subject id
rm_data <- cbind(id = c(1:10), chapter_12_table_1)

# Convert the data from wide to long and convert id, Condition and Angle to factor
rm_data <-  rm_data %>% 
  gather(key = Condition.Angle, 
         value = Reaction_Time, 
         -id) %>%
  separate(col = Condition.Angle, 
           into = c("Condition", "Angle"), 
           sep = -1) %>%
  arrange(id, 
          Condition, 
          Angle) %>%
  convert_as_factor(id, Condition, Angle)

Using the pivot_wider() function

The pivot_wider() function from the dplyr package can also be used accomplish the same task of converting wide to long data. In this case we use the chapter 12 data as the input data frame, select columns Absent0 through Present8 to pivot wider, specify the column names of the selected columns will be placed into a new column called condition.angle, and then the values in to a new column called reaction time. Finally, we will sepatrate the condition and angle variables to be placed in separate columns, condition and angle. The length of the code is roughly the same, but it may be more intuitive to understand the mechanics.

long_data <- 
  pivot_longer(chapter_12_table_1,
             cols = Absent0:Present8,
             names_to = "condition.angle",
             values_to = "reaction_time") %>%
  separate(col = condition.angle,
           into = c("condition", "angle"),
           sep = -1)

References