I would stress that with Excel, changing the underlying data set, especially dimensions of your data is annoying and not very automate-able. This is the case even when the analysis is the exact same or very similar as the original. You either need to copy formulas over for every data set/sheet or write VBA code to automate this, which still isn't as flexible as in R.<p>Excel is better if you want to pore over the data: study a piece, scroll down, study some more. R is better if you can easily pinpoint the areas you think are interesting and index directly into the array. Excel has indexing capabilities (F5 which goes to an address), but not by row/column names.<p>Additionally, you can't do multi-step analysis that involves segmentation/sub-setting of the data very easily with Excel. This is related to the first point about changing an underlying data set or its dimensions. There are ways around this, but they are VERY annoying.<p>I use R for the majority of my analysis. I used to be a Matlab user but needed to find a cheaper tool when I left my job. I still use excel for simple data inspection and when I need to send my analysis to other people to peruse.
The arguments are good and sound, but could be simpler:<p>If you can "grasp" your data by looking at it, use excel.
If your data is too much to look at, use R.<p>Excel lends itself more to shallow exploratory approaches, while with R, you have to think first (or rather: always).
Please don't use spreadsheets for statistical analysis (even the simple kind that the spreadsheet claims to support).<p><a href="http://www.jstatsoft.org/v34/i04/paper" rel="nofollow">http://www.jstatsoft.org/v34/i04/paper</a>
I'm just starting to go beyond Excel, but I chose to learn Python and scipy over R due to reading that R's scripting language didn't match up to Python. Am I handicapping myself by ignoring R?
The arguments are good, assuming you're the only person working on the analysis.<p>As a consultant I'm always constrained to use the greatest common denominator - that is, something that my clients can use, modify and extend <i>themselves</i> after I leave. In 99.9% of the cases, that means Excel, regardless of the task.
So here's a hypothetical question: let's say you have a programmer who knows next to nothing about statistics, knows how to navigate around Excel fairly well, and once allowed himself to be scared away by R.<p>Let's further assume he's suddenly inherited a large volume of load testing data and a mandate to "make something out of it."<p>What's a starting point? I, I mean, <i>he</i> hears about the great visualization stuff in R and understands the importance of it but has no clue where to start.<p>Would HN have any advice for him?
I find R syntax to be really annoying, and I am trying to work more in Octave, a matlab compatible system. I don't like the mostly gratuituous complexity of the R type hierarchy, and prefer to have everything be a matrix. I also detest the use of periods as plain text. FWIW.<p><a href="http://www.octave.org" rel="nofollow">http://www.octave.org</a>
Honestly, once you learn them both, it's best to use Illustrator when you need good presentation and R all other times. I can think of very little that Excel can do that can't be done faster in R and prettier with Illustrator.<p>So I'd condense the argument to: if you're not comfortable with R/Illustrator and don't have time to learn, use Excel.
has anyone used both R and matlab/octave?<p>i've used matlab but not R. For machine learning, I found matlab great because of its great matrix support.<p>what's the pros/cons of R? what advantages does R have over matlab?