hddplot-package {hddplot} | R Documentation |
Cross-validated linear discriminant calculations determine the optimum number of features. Test and training scores from successive cross-validation steps determine, via a principal components calculation, a low-dimensional global space onto which test scores are projected, in order to plot them. Further functions are included for didactic purposes.
Package: | hddplot |
Type: | Package |
Version: | 1.0 |
Date: | 2006-01-09 |
License: | GPL Version 2 or later. |
The most important functions are
cvdisc
: Determine variation in cross-validated accuracy with
number of features
cvscores
: For a specific choice of number of features,
determine scores that can be used for plotting
Note also scoreplot
(plot scores), qqthin
(qqplots, designed
to avoid generating large files when there are many points), and
functions that are intended to illustrate issues that arise
in the plotting of expression array and other high-dimensional data
John Maindonald
Maintainer: John Maindonald <john.maindonald@anu.edu.au>
Maindonald, J.H. and Burden, C.J., 2005. Selection bias in plots of microarray or other data that have been sampled from a high-dimensional space. In R. May and A.J. Roberts, eds., Proceedings of 12th Computational Techniques and Applications Conference CTAC-2004, volume 46, pp. C59–C74.
http://anziamj.austms.org.au/V46/CTAC2004/Main [March 15, 2005].
## Use first 500 rows (expression values) of Golub, for demonstration. data(Golub) data(golubInfo) attach(golubInfo) miniG.BM <- Golub[1:500, BM.PB=="BM"] # 1st 500 rows only cancer.BM <- cancer[BM.PB=="BM"] miniG.cv <- cvdisc(miniG.BM, cl=cancer.BM, nfeatures=1:10, nfold=c(10,4)) miniG.scores <- cvscores(cvlist=miniG.cv, nfeatures=4, cl.other=NULL) subsetB <- (cancer=="allB") & (tissue.mf %in% c("BM:f","BM:m","PB:m")) tissue.mfB <- tissue.mf[subsetB, drop=TRUE] scoreplot(scorelist=miniG.scores, cl.circle=tissue.mfB, circle=tissue.mfB%in%c("BM:f","BM:m"), params=list(circle=list(col=c("cyan","gray"))), prefix="BM samples -") detach(golubInfo) ## Not run: demo(biasedPlots) ## Not run: demo(CVscoreplot)