What is bad(as of now) about R? or What all items R need revision.

vinux

Dark Knight
#1
I know that R is kind of religion for many* of us. I too do marketing job for R in my institute. Few months back I was giving tutorial on R time series and observed few issues with it. Then I dig further on those issues and found those are were known issues. Strange part is, those are in the base packages!!!

First issue is MISTAKES

See AR model output.

Code:
x = arima.sim(list(order=c(1,0,0), ar=.9), n=100) + 50   
mean(x)    
arima(x, order = c(1, 0, 0))
It is showing mean as the intercept.

You could see the details in the link below.
http://www.stat.pitt.edu/stoffer/tsa2/Rissues.htm

This also discuss other time series R issues. Like
Why does arima fit different models for different orders? And what does xreg do?
Why does predict.Arima give strange forecasts?
tsdiag.Arima gives the wrong p-values ...etc.

INEFFICIENT PROCEDURE
This is one issue. Another issue is bad algorithm implementation in the parameter estimation. I don't remember the package name, the package was estimating garch in not in efficient way. fGarch package was giving me the efficient procedure.

INCOMPLETE OUTPUT
Third is minor one. Output improvement. For example, psudo R-square calculation in glm. This would not be a good example for what I meant. Something similar there would be a procedure and we may need some standard measure that is not incorporated.

I see myself to be part of contributing to R in future. I always wanted R in most perfect way. So I thought of knowing limitations of R give better grip on my R sessions/marketing. If you face any similar situation in your subjects, please post here.

I was not looking comment on user friendliness or GUI. I am looking the core part of R. Sorry if the heading looking offensive to any of you.


Regards
Richie
 

Link

Ninja say what!?!
#2
Man...it's so hard for me to say anything bad about R, seeing how I crossed over from SAS. I'm going to think about that for a bit and get back to you.
 

Dason

Ambassador to the humans
#3
I wish there was an easy way to do threading. Mainly so you could have a constant loop running and still grab input from the keyboard through the console.
 

trinker

ggplot2orBust
#7
I think that was r-help they wanted to kill (LINK). That article was pretty harsh and kinda personal I thought. Some valid points but a pretty public place to call people out like that.