DataFrame self-joins · Issue #2996 · pandas-dev/pandas (original) (raw)
Given the following DataFrame
| area | point | test | value |
|---|---|---|---|
| A | 11 | 0 | 1234234 |
| A | 11 | 1 | 12341234 |
| A | 16 | 0 | 234234 |
| A | 16 | 1 | 2343 |
| A | 16 | 2 | 234234 |
| C | 4 | 0 | 234234 |
| C | 4 | 1 | 234234 |
it would be nice if there were a way of grouping say columns area and point and comparing the value per test > 1 with the value for test - 1.
This can be done by iterating over df.groupby(['area', 'point', 'test']) and using the sorting provided by groupby() on the specified columns to compare current and previous values. However, it would be neat if this could also be done in a more Pandas-esque way using something akin to a SQL self-join.
NB request first made in pystatsmodels Google Group; was asked by Wes to create a Github issue for this.