Pairwise operations

[ghost]

The idea is to provide an abstraction that allows series operations that operate between two series to be applied pairwise across an entire data frame, ignoring the columns that have unsupported types for the given operation.

DF.pairwise(df, :correlation) should be similar to https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.corr.html

UPDATE

We decided to go with the same naming from the Series for the operations in the DataFrame.
So we now have a function called correlation/2. The idea is to add more pairwise operations.

• correlation/2 - Pairwise correlation #759
• covariance/2 - Pairwise covariance #763

[josevalim]

For completeness, how does that work in dplyr? Perhaps they would use across? That’s something easy to do with our query today.

[jcmkk3]

For completeness, how does that work in dplyr? Perhaps they would use across? That’s something easy to do with our query today.

In the tidyverse ecosystem, I think that the idiomatic way would be to use the widyr package.