Update: now we have pivot_wider() which works effectively the same. Check it out here.
spread() can reshape dataset containing measures of multiple time points and help query data availability per subject, which can truly be a time saver when there are many subjects with lots of data.
spread(data, key, value)
Now, consider the following (minimal) dataset that has 3 subjects and measures from 3 time points:
## Subject Timepoint Measure
## 1 A T1 10
## 2 A T2 12
## 3 A T3 NA
## 4 B T1 9
## 5 B T2 10
## 6 B T3 9
## 7 C T1 6
## 8 C T2 NA
## 9 C T3 NA
Unless reading line by line, it’s not immediate to tell whether the measure at a certain timepoint is missing or not. It gets even more cumbersome when the list is long. What if say we transform the table to let the Timepoint becomes separate columns, and the values within Measure are stored within those columns instead? This is where spread() can be applied:
dat %>% spread(key = Timepoint, value = Measure)
## Subject T1 T2 T3
## 1 A 10 12 NA
## 2 B 9 10 9
## 3 C 6 NA NA
It gets easier to read along each row to grasp the idea of data availability, plus the dataset becomes much more tidy.