`corr` with ignoring null values

[xuJ14]

Description

With the corrent polars.DataFrame.corr, it returns nulls when columns contain nulls. Sometimes, ignoring nulls could be a wanted option. Pandas has pandas.DataFrame.corr.

[orlp]

Our (newly enforced) policy is that nulls should whenever possible be treated as completely absent by default. The problem with correlation is that you can have a scenario where only one of the two variables is missing.

I think it makes sense to change the default to calculate the correlation only using those rows where both columns have a value. Perhaps a statistician who has a stronger understanding of how the correlation coefficient is used could weigh in on that?

[xuJ14]

I agree with you. The current corr function is the same as check nulls and use your new corr function instead, which is simple to do. So a broader solution would always be welcomed.

[AdrienDart]

Hi, agree with the above, this would be very nice to have the same implementation as pandas to handle nan values as well (https://github.com/pandas-dev/pandas/blob/d928a5cc222be5968b2f1f8a5f8d02977a8d6c2d/pandas/_libs/algos.pyx#L349 => nancorr).