Admittedly, R is very cool.

Quote from zedDoubleNaught:

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.
More...

You can do this in excel. When you "activate" the regression line in the chart, you can check the "show formula" box in the menu. You have to manually type it into the spreadsheet to use it, though.
 
One bad thing, it crashes. I am not even pushing it that hard. Using RODBC to query MSFT SQLServer, ~800 rows * 15 columns, then doing plots.
 
Quote from nitro:

One bad thing, it crashes. I am not even pushing it that hard. Using RODBC to query MSFT SQLServer, ~800 rows * 15 columns, then doing plots.
More...

I have a buddy who runs it on Windows, crashes now and then for him. I'm pushing it pretty hard on Linux and have yet to experience a crash.
 
Quote from johnnyqpublic:

I have a buddy who runs it on Windows, crashes now and then for him. I'm pushing it pretty hard on Linux and have yet to experience a crash.
More...
Hmm, interesting. I may have to try it on Ubuntu and see how it goes...
 
Quote from nitro:

Hmm, interesting. I may have to try it on Ubuntu and see how it goes...
More...

If you can afford to use older libraries (i.e. not the bleeding-edge), debian is a more stable distro and actually the upstream source of the ubuntu project
 
Quote from nitro:

One bad thing, it crashes. I am not even pushing it that hard. Using RODBC to query MSFT SQLServer, ~800 rows * 15 columns, then doing plots.
More...

If that is the total table size I would just read it all into memory.

I have never had it crash and have much larger in-memory data.

I am not using SQL Server though, or any database except flat files.
 
Quote from zedDoubleNaught:

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.
More...

Re the interface, I would suggest StatET with Eclipse. Or RStudio if you do not know Eclipse.

StatET for example has a visual debugger, object inspector, table viewer, graphics management etc......

Fitting a curve to a set of points should be doable in R....
 
Most researchers I know have a love/hate relationship with R. The initial learning curve is steep, the type system is ... sheeesh what where they thinking, etc.

Plenty of great packages, BUT, this can be hit and miss, many times a miss. Quality can be crap (you get what you pay for).

WARNING: If you are using a particular version of R with particular versions of packages, make sure you backup that particular R installer and the packages. Seriously. Believe me, you'll thank me later for this advice when the next version of R doesn't work at all (or as expected) with some packages that have or have not been updated. Unlike Matlab, where things usually work, but we are talking $ when it comes to Matlab.
 
Equalizer, Have you used any packages in R related to neural networks or SVMs? (I recall you said these form part of your trading system)
 
You must log in or register to reply here.
Back
Top