plot.PCA: Plots Principal Component Analysis
In Momocs: Morphometrics using R

Description Usage Arguments Details See Also Examples

View source: R/gr-PCA.R

The Momocs' PCA plotter with morphospaces and many graphical options.

## S3 method for class 'PCA'
plot(x, fac, xax = 1, yax = 2, points = TRUE,
  col = "#000000", pch = 20, cex = 0.5, palette = col_solarized,
  center.origin = FALSE, zoom = 1, bg = par("bg"), grid = TRUE,
  nb.grids = 3, morphospace = TRUE, pos.shp = c("range", "full", "circle",
  "xy", "range_axes", "full_axes")[1], amp.shp = 1, size.shp = 1,
  nb.shp = 12, nr.shp = 6, nc.shp = 5, rotate.shp = 0,
  flipx.shp = FALSE, flipy.shp = FALSE, pts.shp = 60,
  border.shp = col_alpha("#000000", 0.5), lwd.shp = 1,
  col.shp = col_alpha("#000000", 0.95), stars = FALSE, ellipses = FALSE,
  conf.ellipses = 0.5, ellipsesax = TRUE, conf.ellipsesax = c(0.5, 0.9),
  lty.ellipsesax = 1, lwd.ellipsesax = sqrt(2), chull = FALSE,
  chull.lty = 1, chull.filled = FALSE, chull.filled.alpha = 0.92,
  density = FALSE, lev.density = 20, contour = FALSE, lev.contour = 3,
  n.kde2d = 100, delaunay = FALSE, loadings = FALSE,
  labelspoints = FALSE, col.labelspoints = par("fg"),
  cex.labelspoints = 0.6, abbreviate.labelspoints = TRUE,
  labelsgroups = TRUE, cex.labelsgroups = 0.8, rect.labelsgroups = FALSE,
  abbreviate.labelsgroups = FALSE, color.legend = FALSE, axisnames = TRUE,
  axisvar = TRUE, unit = FALSE, eigen = TRUE, rug = TRUE,
  title = substitute(x), box = TRUE, old.par = TRUE, ...)

`x`	an object of class "PCA", typically obtained with PCA
`fac`	name or the column id from the $fac slot, or a formula combining colum names from the $fac slot (cf. examples). A factor or a numeric of the same length can also be passed on the fly.
`xax`	the first PC axis
`yax`	the second PC axis
`points`	logical whether to plot points
`col`	a color for the points (either global, for every level of the fac or for every individual, see examples)
`pch`	a pch for the points (either global, for every level of the fac or for every individual, see examples)
`cex`	the size of the points
`palette`	a palette
`center.origin`	logical whether to center the plot onto the origin
`zoom`	to keep your distances
`bg`	color for the background
`grid`	logical whether to draw a grid
`nb.grids`	and how many of them
`morphospace`	logical whether to add the morphological space
`pos.shp`	passed to pos.shapes, one of `"range", "full", "circle", "xy", "range_axes", "full_axes"`. Or directly a matrix of positions. See pos.shapes
`amp.shp`	amplification factor for shape deformation
`size.shp`	the size of the shapes
`nb.shp`	(pos.shp="circle") the number of shapes on the compass
`nr.shp`	(pos.shp="full" or "range) the number of shapes per row
`nc.shp`	(pos.shp="full" or "range) the number of shapes per column
`rotate.shp`	angle in radians to rotate shapes (if several methods, a vector of angles)
`flipx.shp`	same as above, whether to apply coo_flipx
`flipy.shp`	same as above, whether to apply coo_flipy
`pts.shp`	the number of points fro drawing shapes
`border.shp`	the border color of the shapes
`lwd.shp`	the line width for these shapes
`col.shp`	the color of the shapes
`stars`	logical whether to draw "stars"
`ellipses`	logical whether to draw confidence ellipses
`conf.ellipses`	numeric the quantile for the (bivariate gaussian) confidence ellipses
`ellipsesax`	logical whether to draw ellipse axes
`conf.ellipsesax`	one or more numeric, the quantiles for the (bivariate gaussian) ellipses axes
`lty.ellipsesax`	if yes, the lty with which to draw these axes
`lwd.ellipsesax`	if yes, one or more numeric for the line widths
`chull`	logical whether to draw a convex hull
`chull.lty`	if yes, its linetype
`chull.filled`	logical whether to add filled convex hulls
`chull.filled.alpha`	numeric alpha transparency
`density`	whether to add a 2d density kernel estimation (based on kde2d)
`lev.density`	if yes, the number of levels to plot (through image)
`contour`	whether to add contour lines based on 2d density kernel
`lev.contour`	if yes, the (approximate) number of lines to draw
`n.kde2d`	the number of bins for kde2d, ie the 'smoothness' of density kernel
`delaunay`	logical whether to add a delaunay 'mesh' between points
`loadings`	logical whether to add loadings for every variables
`labelspoints`	if TRUE rownames are used as labels, a colname from $fac can also be passed
`col.labelspoints`	a color for these labels, otherwise inherited from fac
`cex.labelspoints`	a cex for these labels
`abbreviate.labelspoints`	logical whether to abbreviate
`labelsgroups`	logical whether to add labels for groups
`cex.labelsgroups`	ifyes, a numeric for the size of the labels
`rect.labelsgroups`	logical whether to add a rectangle behind groups names
`abbreviate.labelsgroups`	logical, whether to abbreviate group names
`color.legend`	logical whether to add a (cheap) color legend for numeric fac
`axisnames`	logical whether to add PC names
`axisvar`	logical whether to draw the variance they explain
`unit`	logical whether to add plane unit
`eigen`	logical whether to draw a plot of the eigen values
`rug`	logical whether to add rug to margins
`title`	character a name for the plot
`box`	whether to draw a box around the plotting region
`old.par`	whether to restore the old par. Set it to `FALSE` if you want to reuse the graphical window.
`...`	useless here, just to fit the generic plot

Widely inspired by the "layers" philosophy behind graphical functions of the ade4 R package.

plot.LDA

## Not run: 
data(bot)
bot.f <- efourier(bot, 12)
bot.p <- PCA(bot.f)

### Morphospace options
plot(bot.p, pos.shp="full")
plot(bot.p, pos.shp="range")
plot(bot.p, pos.shp="xy")
plot(bot.p, pos.shp="circle")
plot(bot.p, pos.shp="range_axes")
plot(bot.p, pos.shp="full_axes")

plot(bot.p, morpho=FALSE)

### Passing factors to plot.PCA
# 3 equivalent methods
plot(bot.p, "type")
plot(bot.p, 1)
plot(bot.p, ~type)

# let's create a dummy factor of the correct length
# and another added to the $fac with mutate
# and a numeric of the correct length
data(bot)
f <- factor(rep(letters[1:2], 20))
z <- factor(rep(LETTERS[1:2], 20))
bot %<>% mutate(cs=coo_centsize(.), z=z)
bp <- bot %>% efourier %>% PCA
# so bp contains type, cs (numeric) and z; not f
# yet f can be passed on the fly
plot(bp, f)
# numeric fac are allowed
plot(bp, "cs", cex=3, color.legend=TRUE)
# numeric can also be passed on the fly
plot(bp, 1:40, cex=2)
# formula allows combinations of factors
plot(bp, ~type+z)

### other morphometric approaches works the same
# open curves
data(olea)
op <- npoly(olea, 5)
op.p <- PCA(op)
op.p
plot(op.p, ~ domes + var, morpho=TRUE) # use of formula

# landmarks
data(wings)
wp <- fgProcrustes(wings, tol=1e-4)
wpp <- PCA(wp)
wpp
plot(wpp, 1)

# traditionnal measurements
data(flower)
flower %>% PCA %>% plot(1)

# plot.PCA can be used after a PCA
data(iris)
PCA(iris[, 1:4], fac=iris$Species)  %>% plot(1)

### Cosmetic options
# window
plot(bp, 1, zoom=2)
plot(bp, zoom=0.5)
plot(bp, center.origin=FALSE, grid=FALSE)

# ellipses
plot(bp, 1, conf.ellipsesax=2/3)
plot(bp, 1, ellipsesax=FALSE)
plot(bp, 1, ellipsesax=TRUE, ellipses=TRUE)

# stars
plot(bp, 1, stars=TRUE, ellipsesax=FALSE)

# convex hulls
plot(bp, 1, chull=TRUE)
plot(bp, 1, chull.lty=3)

# filled convex hulls
plot(bp, 1, chull.filled=TRUE)
plot(bp, 1, chull.filled.alpha = 0.8, chull.lty =1) # you can omit chull.filled=TRUE

# density kernel
plot(bp, 1, density=TRUE, contour=TRUE, lev.contour=10)

# delaunay
plot(bp, 1, delaunay=TRUE)

# loadings
flower %>% PCA %>% plot(1, loadings=TRUE)

# point/group labelling
plot(bp, 1, labelspoint=TRUE) # see options for abbreviations
plot(bp, 1, labelsgroup=TRUE) # see options for abbreviations

# clean axes, no rug, no border, random title
plot(bp, axisvar=FALSE, axisnames=FALSE, rug=FALSE, box=FALSE, title="random")

# no eigen
plot(bp, eigen=FALSE) # eigen cause troubles to graphical window
# eigen may causes troubles to the graphical window. you can try old.par = TRUE

## End(Not run)