search

sefffal / PairPlots.jl

Beautiful and flexible vizualizations of high dimensional data
https://sefffal.github.io/PairPlots.jl/dev
MIT License
130 stars 7 forks source link

Consider renaming `PairPlots.Correlation` to `PairPlots.PearsonCorrelation` #54

Open juliohm opened 3 months ago

juliohm commented 3 months ago

And adding other types of correlations such as

PairPlots.KendallCorrelation via StatsBase.corkendall PairPlots.SpearmanCorrelation via StatsBase.corspearman

In the future, if https://github.com/JuliaStats/StatsBase.jl/issues/934 gets solved, we can also add PairPlots.ChatterjeeCorrelation.

Also notice that the stable docs contain a typo that says that PairPlots.Correlation() = PairPlots.Calculation(StatsBase.cov)

sefffal commented 1 month ago

Thanks @juliohm for catching these! I agree this should be renamed. Instead of one type per function though, I would like to point out that you can already pass the function of your choice in, eg: PairPlots.Correlation(corspearman). The label in the plot will change accordingly.

As such, Correlation is really not a good name. It will accept any function of two variables.

I would welcome suggestions for what to name this instead. Something like PairwiseStat?

juliohm commented 1 month ago

Thank you @sefffal for taking a look into it. I believe that PairPlots.Pairwise(fun) would be clear in this context? Do you plan to restrict the functionality to stats functions, or should it work with any pairwise function?