DOC: Flesh out the R comparison section of docs

[hayd]

I guess quite a lot of people come from an R background, and perhaps a good material would be a conversion table for pandas vs R functions/idioms etc. in http://pandas.pydata.org/pandas-docs/dev/comparison_with_r.html

Perhaps this site could offer some functions to consider including: http://www.statmethods.net/management/variables.html

[jreback]

I guess R is famous for obfuscation (of syntax)?

[hayd]

I suspect it'll be a many-to-one table :)

[cpcloud]

ha! u guys are funny. what the heck is attach? is like attach(x) == globals()['x'] = x?

[jreback]

isn't R intuitive?

[cpcloud]

where the heck are cyl and vs coming from? This

attach(mtcars)
aggdata <- aggregate(mtcars, by=list(cyl,vs), FUN=mean, na.rm=TRUE)
detach(mtcars)

works only if you do the attach(mtcars)? wtf are the scoping rules in R? no such thing exists in Python without a lot of magic...

[jtratner]

Attach basically is like saying 'make all of the columns of the data frame global variables'

[jtratner]

It has a companion method detach. I think there's also a with - like statement that scopes just to the function call. Have you seen the model syntax yet? a ~ b I totally get that it's useful, but it's a little unsettling when you are used to being able to explicitly trace all names in the document.

[cpcloud]

patsy + statsmodels + pandas >>>>> R

[cpcloud]

magic regarding scope and namespaces :-1:

[cpcloud]

anyway comparisons are useful to show people how awesome pandas is :)

[hayd]

related http://stackoverflow.com/questions/17621325/equivalent-pandas-function-to-this-r-aggregation

Anyone fancy spamming the pandas/R/.. mailing lists to see if anyone is interested in doing this?

[hayd]

http://stackoverflow.com/questions/18005305/implementing-r-scale-function-in-pandas-in-python

[TomAugspurger]

http://stackoverflow.com/questions/19237878/subsetting-a-python-dataframe

[hayd]

https://groups.google.com/forum/#!topic/pydata/1eNURQsflNw

A while back I started making some notes on how to do the various recipes in O'Reilly's R Cookbook (http://shop.oreilly.com/product/9780596809164.do) with Numpy, Pandas, Scipy.

I haven't had time to complete it so I'm sharing it in it's current state, and trying to get some community help to fill in the gaps.

I think this could be an extremely useful resource to encourage and help transition lots of people from R to Pandas.

So here's the notes:

http://notes.lexual.com/tech/r_numpy_pandas_cookbook.html

And here's the github repo, patches more than welcome!

https://github.com/lexual/sphinx-notes/blob/master/source/tech/r_numpy_pandas_cookbook.rst

Cheers,

Lex.

These look useful, shame there are some sections which are XXX-titled, as would be nice to have a todo list on this for areas to flesh out.

[8bit-pixies]

I know that this section is more for pandas vs R, but I'm wondering would it be worthwhile to place some of the R functions if it isn't really related to pandas, for example: aaply, alply, or maybe dlply?

[jreback]

@chappers

want to add this: http://stackoverflow.com/questions/20905713/equivalent-of-rs-tapply-in-python-pandas

[8bit-pixies]

hmm, would you want this to go under the reshape/cast section, or in the with section, since it could be done in R using dcast as well:

mydf <- data.frame(
  Animal = c('Animal1', 'Animal2', 'Animal3', 'Animal2', 'Animal1', 'Animal2', 'Animal3'),
  FeedType = c('A', 'B', 'A', 'A', 'B', 'B', 'A'),
  Amount = c(10, 7, 4, 2, 5, 6, 2)
)

# Stackoverflow example
with(mydf, tapply(Amount, list(Animal, FeedType), sum))

# Using reshape
require(reshape2)
dcast(mydf, Animal ~ FeedType, sum, fill=NaN)

In either case the solution would be whats in Stackoverflow (and very similar to the solution in the reshape/cast section of the current docs).

[jreback]

you can out under the more common / useful and put a link / statement in the other (as they r in the same page)

read it as if you are an R user doing the most common operation (eg what is normally recommended to R Users) and you want to convert to pandas

[jreback]

there are of course similar cases in pandas where multiple solutions present (eg imagine a vectorized function vs using apply)

one solution maybe faster or simpler or they may both be appropriate

[jreback]

think this is closable after the multiple PR's by @chappers