Admittedly, R is very cool.

Discussion in 'Automated Trading' started by nitro, Oct 9, 2010.

  1. nitro

    nitro

    I have been using R for the first time in a serious manner, and it is extremely well done and useful.

    I decided to study some data. First I built a C# program that downloads data from the web, parses the XML into objects, uses an ORM to connect to a SQL database, and stores the parsed data in tables.

    Then I use R with RODBC to connect to the SQLServer database/table using specialized queries to do all sorts of interesting analysis on the data. The graphics through plots that understand results in table/matrix form is awesome.

    I knew almost nothing about R, and using this book

    http://www.amazon.com/R-Nutshell-De...170X/ref=sr_1_1?ie=UTF8&qid=1286657704&sr=8-1

    I am able to do some interesting analysis by following the examples.

    Major Kudos. Now, if I were just a statistician...
     
  2. Does it have an interface or GUI front end, or is it all command line? I've been using QTOctave on my Linux machine, it has a GUI-command line mix. It was not too painful to start out graphing some data tables as mesh plots.

    I'd be interested in learning R, it sounds like the standard for data number crunching. However, if the learning curve is too steep, probably never will.

    One feature I'm looking for is a package or program that will give a formula from a set of data points. If it can do it in 3 dimensions, would be even better. I don't know if QTOctave can do it, haven't had time to investigate it to that extent.
     
  3. Huh?
     
  4. I mean, suppose you have a graph, and you can see the data points are close to a line or a curve, it will give you the formula for the line or curve that best matches it. For example, in Excel, I can plot a scatter diagram and do a linear regression to draw the best fitting line through the points. In the Excel chart, I can see the line; but then I want the app to give me the formula for that line, instead of myself having to figure out the slope and y-axis intercept and figure out the formula.
     
  5. Using JGR and the Deducer library can get you something GUI-ish. Depends on what you're looking for, I guess. Example here: http://www.youtube.com/watch?v=PE5mgNbq7xQ

    If you're a programmer, the learning curve isn't that steep. If not, you will have to learn to program AND program in R.

    You may be talking about a simple linear regression, though I'm guessing not. Either way, R probably has a function or library that will do what you want. CRAN (http://cran.r-project.org/) is a great resource for finding libraries.

    Also check out Eureqa (http://ccsl.mae.cornell.edu/eureqa) for doing equation searches.
     
  6. Thanks, this will help. Mostly, I'm hoping for getting the curvilinear relationship formulas, or for data sets with 3 dimensions. Straight lines, I can usually figure out on my own, but the curvilinear ones, or 3-d formulas, is a bit beyond my math which has gotten a bit rusty over the years.
     
  7. There are many GUI-for-R projects, but most users believe they're on a road to nowhere. Command-line R doesn't have a 'steep learning curve' -- I'm not sure how that idea took hold. Spend a couple hours reading the short, clear Smith/Venables book that's included with the standard R distribution http://cran.r-project.org/doc/manuals/R-intro.pdf before claiming R is difficult. Re: Eureqa, yes, good suggestion.
     
  8. WOW !!!!!!!!!!!!! In the last 5 minutes I've downloaded Eureqa and put in some simple data, and that's the type of tool I was looking for !!

    Thank you johnnyqpublic, this is AWESOME !!! I probably never would have found that on my own !! I can't say it enough, this is a great help to me, I really appreciate the suggestion.

    :)
     
  9. You're welcome! :)

    I had the same reaction when I stumbled across it, I think it was on the programming Reddit or something...
     
    #10     Oct 9, 2010