data-science
data-science copied to clipboard
n3mo
Please add to racket package repository
Thanks for the recommendation! This has been on my to-do list for some time. What keeps me from sharing to date is the rather unfinished state of the project. At some point I suppose it could become useful to others though, even if in my own mind it feels unpolished and incomplete. I'll try to make it an official racket package soon!
Thanks - making it available as a package might attract collaborators (if that is something you are interested in)
Hi, n3mo! Are you considering packaging your great data-science module? I am interested in using it.
Definitely still planning to do so, though I'm clearly taking my time in deliberating. Essentially what's keeping me sharing this are some design questions. Essentially I worry about making it available before I fully commit to how data should be represented. I'd hate to promote the package, then realize a switch is in order, and then have to worry about supporting bad decisions down the road. In short, my question is:
Should data be represented in a lisp-y way (lists and vectors, essentially), or should I adopt something like Alex Harsanyi's Data Frame package to make things more reminiscent of R's data frames and python's pandas? So far I've been building things on top of lisp-y data structures. But my work in R reminds me daily of the value of data tables in statistical work. I'd welcome any feedback on your thoughts regarding these concerns.
Thank you for asking my opinion on data-science! I would love if data-science were able to interact with other packages like plot and data-frame. Using the data structures offered by data-frame would make data-science more akin to R and in my view this is a good thing. A word on how I would use data-frame: I would use it as a replacement for R, doing statistical calculations and plots all in Racket instead of R. I am an occasional R user, not a professional statistician.