I have a large 3D data set (~100 MBs in size) and good old Gnuplot does not only take too long to produce a plot, but I noticed that it produces inconsistent plots when the display grid is changed. I tried Mathematica and to my disappointment it didn't even produce the figure. Then I tried R, which seems much faster than Gnuplot, but I couldn't find a good reference on how to set the 3D plot parameters. Any helpful suggestions?
maybe it is feasible to just plot a random subsample of the data within the environment you know.
in R, pretty much everything is possible. it is often helpful to build a minimal example and ask at stackoverflow or r-help mailing list about errors or solutions.
hi Mohamed
R has at least 3 packages providing 3D plots.
package: scatterplot3d, function: scatterplot3d, manual: http://cran.r-project.org/web/packages/scatterplot3d/scatterplot3d.pdf
package: rgl, function: plot3d, manual: http://cran.r-project.org/web/packages/rgl/rgl.pdf
and package: Rcmdr, function: scatter3d, manual: http://cran.r-project.org/web/packages/Rcmdr/Rcmdr.pdf
best regards
Your data set may be large enough for you to benefit from data reduction techniques so you can make out the forest and not just the trees. Approaches include various smoothing methods (e.g. bin-averaging to a grid, or spline fitting) and aggregated-graph methods. It is hard to say much more because you've not said the kind of graph you want. If you have *f=f(x,y)* data, I suggest dropping the idea of 3D plots and using instead image plots, contour plots, etc., at least for the publication phase, because 3D plots are hard for the viewer to interpret and image/contour/etc plots are unequivocal.
The R is true doesn't have a lot of publications on 3D data. You can actually try Matlab which is a more advanced software for 3D data and works that are intended to display 3D graphs and other outputs. You may need to read a little on the Matlab though as it is highly programming, unlike the R, you may need to buy it to be able to use use it for analysis of large data such as yours. You can get the Matlab free downloads from http://www.mathworks.com/downloads/ and you may find a lot of books on it from the same site. Read this for the graph aspect http://asa.aip.org/figures.pdf and get this paper from http://hpc.sagepub.com/content/23/3/277.abstract and let see if it helps.
Good Luck
Thanks to all of you who responded to my query. By 3D plots I basically meant surface plots. I am not a Matlab user, but I doubt it can handle a data size that Mathematica couldn't. For now, I am getting by an option in Gnuplot that allows skipping of every n data points. (I found this easier than finding the right grid display commands in R.) My data sets are numerical results that consist of portions with rapid changes (which you want to capture) and slow-varying portions. The aim is to display the rapid changes as much as necessary. I am interested in following Dan's suggestions, but at the same time wonder if these data smoothing methods would make the plotting much more time consuming.
Thanks Patrice. No, I haven't tried it. Do you think it solves the problem I discussed above?
Paraview should be able to handle a 100 Mb file. Paraview is free, multiplateform and easy use.
- Badges
- Science topic