Function to Plot Principal Component Analysis Loadings and Scores
Function to display the results of a Principal Components Analysis (PCA) from the saved object from gx.mva, gx.mva.closed, gx.robmva, gx.robmva.closed or gx.rotate as biplots. Various options for displaying loadings and scores are available, see Details below.
gx.rqpca.plot(save, v1 = 1, v2 = 2, rplot = TRUE, qplot = TRUE, rowids = NULL, ifrot = TRUE, main = "", cex.lab = 0.9, cex.main = 0.9, rcex = 1, qcex = 0.8, rcol = 1, qcol = 1, ifray = TRUE, if34 = TRUE, ...)
save |
a saved object from the execution of function |
v1 |
the component to be plotted on the x-axis of the biplot, default is the first component, |
v2 |
the component to be plotted on the y-axis of the biplot, default is the second component, |
rplot |
the default is to plot the variables. If the variables are not required set |
qplot |
the default is to plot the observation (individual, case or sample) scores. If scores are not required set |
rowids |
‘switch’ to determine if the input matrix row numbers are to be displayed instead of default plotting symbols. The default is for default plotting symbols, i.e. |
ifrot |
by default the post-Varimax rotation scores are displayed if a rotation has been made, see |
main |
an alternate plot title from that generated automatically from information in the saved object, see Details below. |
cex.lab |
the text scale expansion factor for the axis labels of the display, by default |
cex.main |
the text scale expansion factor for the display title, by default |
rcex |
the text scale expansion factor for the variable names in the display, by default |
qcex |
the text scale expansion factor for the observation symbols, row numbers or sample IDs in the display, by default |
rcol |
the colour of the variable names in the display, by default |
qcol |
the colour of the observation symbols, row numbers or sample IDs in the display, by default |
ifray |
by default ‘rays’ are plotted from the R loadings origin to the variable name in the same colour as the variable names, to omit the ‘rays’ set |
if34 |
this applies to a robust biplot only, by default the top (3) and right side (4) of the plot frame are annotated in black with the R loadings scale and with axis titles in the same colour as the variable names. To suppress these annotations set |
... |
further arguments to be passed to methods concerning the plot. |
If main is undefined the name of the matrix object supplied to the function is displayed in the plot title. On the line below the name of the data matrix from which the PCA was derived is displayed. However, if an alternate plot title is preferred it may be defined, e.g., main = "Plot Title Text". If no plot title is required set main = " ".
If the variable names are longer than three characters the display can easily become cluttered. In which case the user should redefine the variable names in the input matrix from which the PCA was derived using the dimnames(matrix.name)[[2]] construct, and run the generating function again. Alternately, the variable names in the saved object may be changed directly via a redefinition of save$matnames[[2]].
Information on the percentage of the variability explained by each component, and whether or not rotation has been undertaken, is recovered from the saved object and used to appropriately label the plot axes. Note that for non-robust models the percentage variability explained will be the same as the percentage variability explained by the corresponding eigenvalues.
The following describes the available plot option combinations, the first being the default:
rplot = TRUE & qplot = TRUE & rowids = NULL, crosses (pch default) and variable names
rplot = TRUE & qplot = FALSE & rowids = NULL, variable names only
rplot = FALSE & qplot = TRUE & rowids = NULL, crosses (pch default) only
rplot = FALSE & qplot = TRUE & rowids = TRUE, input matrix row numbers only
rplot = FALSE & qplot = TRUE & rowids = FALSE, input matrix row identifiers
rplot = TRUE & qplot = TRUE & rowids = TRUE, input matrix row numbers and variable names
Because functions gx.mva, gx.mva.closed, gx.robmva or gx.robmva.closed require a matrix as input the sample IDs that may be in a data frame may be lost. They may be re-inserted by copying dimnames(...)[[1]] from a data frame into the matrix. Alternately, to plot in the component score space with Sample IDs, the scores can be recovered from the saved object, e.g., save$rqscore[, 1] and save$rqscore[, 2], and used as the x- and y-coordinates in function xyplot.tags with the sample IDs from the source data frame. Appropriate plot and axis titling can be displayed by setting the function arguments ‘by hand’.
The scaling for the PCA ensures that the loadings and scores may be plotted in the same numerical space to create a biplot. However, when robust covariance matrix and mean estimation is employed this is not possible due to the extreme values some scores may attain for outliers. In this case the variables are plotted in the same physical space but with different scaling and addition scaling provided for the top and right side of the plot frame.
Robert G. Garrett
Reimann, C., Filzmoser, P., Garrett, R. and Dutter, R., 2008. Statistical Data Analysis Explained: Applied Environmental Statistics with R. John Wiley & Sons, Ltd., 362 p.
Venables, W.N. and Ripley, B.D., 2001. Modern Applied Statistics with S-Plus, 3rd Edition, Springer, 501 p.
## Make test data available data(sind) data(sind.mat2open) attach(sind) ## Save PCA results and display biplots sind.save <- gx.mva.closed(sind.mat2open) gx.rqpca.plot(sind.save) gx.rqpca.plot(sind.save, main = "Howarth & Sinding Larsen Stream Sediments\nclr transform", pch = 4, rcol =2, qcol = 4) gx.rqpca.plot(sind.save, rplot = TRUE, qplot = FALSE, rowids = NULL) gx.rqpca.plot(sind.save, rplot = FALSE, qplot = TRUE, rowids = NULL) gx.rqpca.plot(sind.save, rplot = FALSE, qplot = TRUE, rowids = TRUE) gx.rqpca.plot(sind.save, rplot = TRUE, qplot = TRUE, rowids = FALSE, rcol = 2, qcol = 4) ## ## Alternately xyplot.tags(sind.save$rqscore[, 1],sind.save$rqscore[, 2], ID, cex = 0.9) ## Clean-up and detach test data rm(sind.save) detach(sind)
Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.