travels in 3d space

12
Travels in 3D space Data ellipsoids, biplots, and rgl movies SCS Seminar, Oct., 2008

Upload: zofia

Post on 03-Feb-2016

50 views

Category:

Documents


0 download

DESCRIPTION

Travels in 3D space. Data ellipsoids, biplots, and rgl movies SCS Seminar, Oct., 2008. Prelude. Multivariate data often needs > 2D Usually reduced to multiple 2D views Static 3D visualization often Badly rendered Lacks control of perspective Lacks direct manipulation of viewpoint - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Travels in 3D space

Travels in 3D space

Data ellipsoids, biplots, and rgl movies

SCS Seminar, Oct., 2008

Page 2: Travels in 3D space

Prelude

• Multivariate data often needs > 2D Usually reduced to multiple 2D views

• Static 3D visualization often Badly rendered Lacks control of perspective Lacks direct manipulation of viewpoint

• Dynamic 3D visualization ggobi / rggobi – powerful dynamic graphics, but

crummy rendering rgl – beautiful rendering, good interactive control of

perspective, viewpoint, etc., weak 3D “tours”• Goal: Explore 3D vis & animation with rgl

Page 3: Travels in 3D space

Iris data: rgl:::plot3d()

data(iris); library(rgl)col <-c("blue", "green", "red")[iris$Species]plot3d(iris, type="s", size=0.4, col=col, cex=2, box=FALSE, aspect="iso")

Page 4: Travels in 3D space

Add data ellipse

cov <- cov(iris); mu <- mean(iris)plot3d( ellipse3d(cov, centre=mu, level=0.68), col="gray", alpha=0.2, add = TRUE)

Page 5: Travels in 3D space

Add PC axes

source("c:/R/functions/ellipse3d.axes.R")axes <- ellipse3d.axes(cov, centre=mu, level=0.72, labels=TRUE)M1 <- par3d("userMatrix")

Page 6: Travels in 3D space

Rotate to show PC1 & PC2: Biplot view

# hand rotate / zoom, then save current positionM2 <- par3d("userMatrix")

Page 7: Travels in 3D space

Rotate to show PC1 & PC3

M3 <- par3d("userMatrix")

Page 8: Travels in 3D space

Rotate to show PC2 & PC3

M4 <- par3d("userMatrix")

Page 9: Travels in 3D space

Biplot movie: rotation to PC coordinates

interp <-par3dinterp( userMatrix=list(M1, M2), extrapolate="constant", method="linear")movie3d(interp, duration=4, fps=8, movie="biplot3d-iris")

Page 10: Travels in 3D space

Grand tour: Interpolation thru multiple views

interp3 <-par3dinterp(userMatrix=list(M1, M2, M3, M4, M3, M2, M1), extrapolate="constant", method="linear" )movie3d(interp3, duration=6, fps=8, movie="biplot3d-iris3", dir="../anim")

Page 11: Travels in 3D space

View in PCA space: bpca package

library(bpca)col <-c("blue","green","red")[iris$Species] plot(bpca(iris[-5], method='hj', lambda.end=3), rgl.use=TRUE, obj.cex=.8, obj.col=col, simple.axes=FALSE, box=TRUE)

movie3d(spin3d(), movie="biplot3d-bpca“)

movie3d requires ImageMagik:

http://www.imagemagik.org

Page 12: Travels in 3D space

movie3d details

• rgl can only write .png files. movie3d() just writes a lot of them• Default: writes duration x fps .png files to the current or dir=

directory Need to figure out how time relates to angular rotation E.g., spin3d() uses rpm=5, so need duration=12 to get 360o

• Calls: convert –delay 1x{fps} movie*.png movie.gif, producing an animated .gif image. Animated .gif can be viewed as an animation in most browsers (e.g.,

Firefox) and Powerpoint, but will not animate in many image viewers (IM display, …) or in .pdf files created from PPT (like this one).

Once the frames have been created, you can run convert in a DOS shell with other parameters, or create other movie formats, e.g., mpeg, pdf, ...

• LaTeX beamer: I’ve discovered how to run movie3d .gif animations from a beamer .pdf file. The room left on this slide is too small to say how.