Hacker News Reader

Comment by tomrod | Hacker News Reader

Unless you need the frontier of statistical methods and are unable to code them yourself, I highly recommend not engaging R.

While Hadley Wickham has done amazing things to make the R language actually useful, Python and Julia are better for data applications. Also, bindings to the R language exist, obviating the need to be tied into it completely. Thus, even if you need the frontier, you have it available to you.

View on HN

R is first and foremost an interactive language, and the focus of the libraries, tooling, and community reflects that.

I believe working in pandas/sklearn/jupyter ecosystem is much slower (>50%) if you are doing EDA and statistical modelling (ML) than tidyverse/tidymodels/rstudio. The exception is deep learning and adjacent fields (like computer vision).

Of course Python is better at everything else, so adding an extra languages might or might be worth the hassle.

Just as a followup, when asked I also excoriate Python for its gap in typing for data applications. But that gap isn't as wide as the gulf between Python & Julia vs R

I would say R has a beautiful functional approach which is very hard to replicate in python. It is also faster than Python and Julia for tabular/dataframe data. Ggplot is fantastic - I tried to recreate some relatively simple plots (scatterplot with a LOESS line of best fit coloured by group) and failed to do this in Python easily.

Reply on news.ycombinator.com