Wide or Long
When do we encounter wide or long formats ?
Wide and long formats are generic terms to express how data are stored when you have repeated measurement for one observed unit. You will encounter these formats when using panel or quasi-panel data, i.e. time series, choice experiments, etc.
Wide format
With a wide format, a subject’s repeated responses will be in a single row, and each response is in a separate column.
In the following table, the respondents responded three times to a question. The answers are stored in the three columns (from ans_1 to ans_3).
All the other columns correspond to unique characteristics (or answers) of the respondents. They are not repeated.
## # A tibble: 5 x 6
## resp_id age city ans_1 ans_2 ans_3
## <int> <int> <chr> <int> <int> <int>
## 1 1 43 Cape Town 2 1 1
## 2 2 48 Pretoria NA 2 2
## 3 3 45 Cape Town 3 3 2
## 4 4 40 Pretoria 2 1 2
## 5 5 31 Cape Town 2 2 1
Long format
With a long format, each row corresponds to one response. So each respondent will have data in multiple rows.
## # A tibble: 15 x 5
## resp_id age city Question Answer
## <int> <int> <chr> <int> <int>
## 1 1 43 Cape Town 1 2
## 2 1 43 Cape Town 2 1
## 3 1 43 Cape Town 3 1
## 4 2 48 Pretoria 1 NA
## 5 2 48 Pretoria 2 2
## 6 2 48 Pretoria 3 2
## 7 3 45 Cape Town 1 3
## 8 3 45 Cape Town 2 3
## 9 3 45 Cape Town 3 2
## 10 4 40 Pretoria 1 2
## 11 4 40 Pretoria 2 1
## 12 4 40 Pretoria 3 2
## 13 5 31 Cape Town 1 2
## 14 5 31 Cape Town 2 2
## 15 5 31 Cape Town 3 1
As you can see, each respondents are described in three rows, one per question. The question is described by a question id, and an answer.
Note also variables that do not change across responses will have the same value in all the rows of the corresponding respondent. This leads to some repetitions, but may be useful when conducting data analysis.
Depending on data analysis needs, you will have to change from one format to another. Therefore, it is important that you master the functions to reshape data in one format or another.