Open hadley opened 8 years ago
Thanks! I'll have a read, looks interesting. :)
I've been thinking about this a bit and I'd be interested in your thoughts, Hadley.
While I think it's good to incorporate some row+col reordering features into visdat
, I that reordering could potentially be a useful "forensic/investigatory" process in data cleaning, where you try and work out patterns in data entry or missing data.
So my concerns are that in introducing these features like seriation or expectations (as in #15) into visdat
, it might get away from the idea of one package doing one thing simply, and perhaps it would be best to make another package that uses seriation / expectations to explore data features, and potentially missing data patterns altogether.
I think it's reasonable to include both original and seriated orders - I don't see any reason to privilege the original order of the rows/columns over potentially other more useful orders.
That's a good way of looking at it. Thanks for that mate, very much appreciated.
just a note to myself to keep an eye on this issue here: https://github.com/dgrtwo/widyr/issues/4 as this looks like it will be very very handy
https://cran.r-project.org/web/packages/seriation/vignettes/seriation.pdf