Excel versus R

[axiomOfChoice]

Dear users,

I was recently in an ongoing debate with one of my colleagues about the merits of Excel, versus the merits of R. I recently wrote a blog article on the subject, which has become something of a talking point in the office: http://forgetfulfunctor.blogspot.com/

I come from a maths/programming background, and have only become acquainted with Excel through the workplace. I apologise in advance for any offense taken about the average Excel user: clearly I need to get the input of professionals. I am interested in what professional Excel users (financial modelers and the like) think about its weaknesses, strengths, and the potential for alternatives, such as R, to replace it.

I would appreciate any feedback on the article, and am trying to get input for a follow up. All sensible comments will be happily received.

 

[Cindy Ellis]

I've used both Excel and R to solve statistical problems. If you've done the same, I'm not telling you anything you don't already know, but for what it's worth, here's my comparison:
R borders on user-hostile, as the counterpoint to the user friendliness of Excel.
R's help system is not very helpful. That's true of much of Excel's help system also.
R is very powerful, but requires writing a script (or remembering the script-based commands) to do anything worth doing. Once you've created and executed a script, it's hard to figure out if the output actually provided an answer to the question you meant to ask, or if it just spit out a meaningless number. This meant I went back to Excel to validate any answer I got in R...which meant I used Excel the next time I needed to do whatever it was I was trying to figure out in R.
Excel is also very powerful, but provides substantial non-statistical functionality and limited statistical functionality without the need to write a program. It can be used as a low end database and a high-end calculator all without writing a line of code. For those of us willing to write code, Excel can create completely customized applications.
There are some things that R does really well that Excel just can't touch (as far as I know). R's ability to plot a function, including probability distributions, without creating or entering any data for the function is very cool, and sometimes necessary. I also don't remember finding gamma or beta functions in Excel, but they may be there. That's something I'll never use in real life, but was essential to (barely) passing my mathematical statistics class.

I haven't used R in about 2 years now, but I still remember that it had lots of neat features that I would never use. I use Excel all the time...and it also has lots of neat features that I will never use.

I do use Excel for basic statistical calculations, but if I need hypothesis tests more complicated than a paired t-test, my current statistical tool of choice is MiniTab. My preferred statistical tool is SPSS (I used to write macros for that, too), but my company hasn't invested in it, and as you mentioned, these programs can be prohibitively expensive.

Cindy

 

[axiomOfChoice]

Hi Cindy,

You make some excellent points, and I think that it points out that there is a sound place for Excel in the analysts toolkit. FYI, this came up as part of a staff training program, and I was trying to make the point that focusing solely on Excel could deskill our employees.

Thanks for your input.

 

[Cindy Ellis]

Any software program (or even a calculator) that simplifies what we have to think about can deskill us. The biggest problem with any statistical or analytical software is that it removes the need to understand what you want to do before you do it. It's really easy to use a program to churn out page after page of entirely meaningless statistics without any understanding of the problem at hand. The output, however right or wrong it might be, is essentially useless if it can't be interpreted properly.
On the positive side, though, the use of a program frees us from tedious calculations, and assuming the input is correct (including the correct choice of equations, statistics, confidence levels, etc.) the output is likely to be correct within the limitations of digital computing, with a few notable bugs here and there.

 

[Norie]

Isn't there an R package/add-in for Excel?

 

[Gerald Higgins]

I am not a statistician, and I had never heard of R before reading this thread. I was interested to read Cindy's posts, and just wanted to offer a small correction and add to the capabilities of Excel.

R's ability to plot a function, including probability distributions, without creating or entering any data for the function is very cool

Although I've never done it myself, I think it can be done in Excel, like here
http://www.excelhero.com/cgi-bin/mt/mt-search.cgi?blog_id=4&tag=Advanced%20Charting&limit=20

Apologies if this is different to what Cindy was referring to.

 

[SydneyGeek]

Stephen Bullen created a really cool worksheet for plotting polynomials. http://oaltd.co.uk/Excel/Default.htm and download ChtFrmla.zip

A stats package that I used years ago was called GraphPad Prism. It was fairly user-friendly but what I liked about it was (1) the graphing package and (2) the fact that the guy who wrote it, produced a really detailed booklet on why you were running a particular analysis. Think his name was Harvey Motulsky.

Denis