This page is for features that people would like to see in the R language, or in current or new packages.
Note that it is almost certainly futile to wish for changes in behavior of basic R functions, because much (actually an unknown quantity of) software depends upon the current behavior of basic R functions, and the authors of R have no wish to break existing software. For example, although many people might wish that “$” on lists did not do partial matching of names, there is an unknown amount of software that depends on this behavior and there is virtually no chance that the current behavior will be changed.
Add new wishes here (already added are: NEWS, svn log):
Wouldn’t it be nice if
boxplot() had an option to drop empty groups? Plotting nested data is typical of the problem. A drop.empty=F default would not compromise backward compatibility.
Optimizer:
http://www.ingber.com/#ASA
A very fast global optimizer with non-linear cost function and non-linear constraints, specifically ASA
Business Cycle analysis tools: “HP, Baxter-King and Christiano-Fitzgerald filters, the Bry-Boschan-Harding-Pagan procedure for the datation of turning points” as described in detail in Chapter 15 of the Grocer manual. Grocer is the open source Econometrics add-on to SciLab here:
http://dubois.ensae.net/grocer_manual_v1.1.zip . I’ve done some work to generate the Baxter King cycle via gretl here:
http://www.itbizvision.com/ion/tiki-index.php?page=BaxterKing but what is needed is much faster C or FORTRAN functions for all these filters and turning point datation and prediction methods.
Implement correlogram (graphical display of correlation coefficients) ideas from
http://www.math.yorku.ca/SCS/sasmac/corrgram.html - the corresponding
paper. Perhaps package
ellipse could be a nice place for this as it already provides ellipses. Maybe separate package with ellipse as dependency would be even better.
Kevin Wright created the
corrgram package (on CRAN) that provides this feature.
Wouldn’t it be nice to be able to zoom and pan R graphics device windows? Like MATLAB? And Python via Matplotlib? The iplots package seems stalled at v. 1.07, the award winning JGR front end balks at large data sets and (as I recall) does not work for all plot types. And simple fixes using
locator() are quite limited. Can’t there just be DEVICE definitions that include zoom and pan controls on plot borders? R ships now with Tcl. If needed for more flexible graphics, couldn’t it also ship with Java? I believe R would disseminate much more widely with this enhancement.
Wouldn’t it be very nice to be able to reliably stop R when it goes into some hugely lengthy computation, instead of having the operating stop it and lose all data in memory and waste much user time re-creating the data? I know ESC is supposed to do that but it does not work most of the time. A much more reliable mechanism is necessary.
Wouldn’t it be nice if the
arima(,1,)(,,) function was debugged and worked and forecasted correctly as clearly explained here:
http://www.stat.pitt.edu/stoffer/tsa2/Rissues.htm (which explains what corrections are necessary and how to do them)? Or is there a valid reason ARIMA is not debugged and is there another side to the story?
R Wiki powered by
and optimized for Firefox
