Panel principal components analysis.
Syntax
group_name.(options) [indices]
where the elements to display in loadings, scores, and biplot graph form (“out=loadings”, “out=scores” or “out=biplot”) are given by the optional indices, (e.g., “1 2 3” or “2 3”). If indices is not provided, the first two elements will be displayed.
Basic Options
out=arg (default=“table”) | Output type: eigenvector/eigenvalue table (“table”), eigenvalues graph (“graph”), loadings graph (“loadings”), scores graph (“scores”), biplot (“biplot”). |
eigval=vec_name | Specify name of vector to hold the saved the eigenvalues in workfile. |
eigvec=mat_name | Specify name of matrix to hold the save the eigenvectors in workfile. |
prompt | Force the dialog to appear from within a program. |
p | Print results. |
Table and Eigenvalues Plot Options
The number of elements to display in the table and eigenvalue graph form is given by the minimum of the elements specified using the “n=”, “mineigen=” and “cproport=” options.
The default eigenvalue graph shows a scree plot of the ordered eigenvalues. You may use the “scree”, “cproport”, and “diff” option keywords to display any combination of the scree plot, cumulative eigenvalue proportions plot, or eigenvalue difference plot.
n=arg (default=all) | Maximum number of components. |
mineigen=arg (default=0) | Minimum eigenvalue. |
cproport=arg (default=1.0) | Cumulative proportion of eigenvalue total to attain. |
scree | Display a scree plot of the eigenvalues (if “output=graph). |
diff | Display a graph of the eigenvalue differences (if “output=graph). |
cproport | Display a graph of the cumulative proportions (if “output=graph). |
Loadings, Scores, Biplot Graph Options
scale=arg, (default= “normload”) | Diagonal matrix scaling of the loadings and the scores: normalize loadings (“normload”), normalize scores (“normscores”), symmetric weighting (“symmetric”), user-specified (arg=number). |
cpnorm | Compute the normalization for the scores so that cross-products match the target (by default, EViews chooses a normalization scale so that the moments of the scores match the target). |
nocenter | Do not center the elements in the graph. |
mult=arg (default=”first”) | Multiple graph options: first versus remainder (“first”), pairwise (“pair”), all pairs arrayed in lower triangle (“lt”) |
labels=arg (default=“outlier”) | Scores label options: identify outliers only (“outlier”), all points (“all”), none (“none”). |
labelprob=arg (default=0.1) | Outlier label probability (if “labels=outlier”). |
autoscale=arg (default=1.0) | Rescaling factor for auto-scaling. |
userscale=arg | User-specified scaling. |
Covariance Options
period | Compute period (within cross-section) panel covariances and related statistics. The default is to compute contemporaneous (between cross-section) measures. |
cov=arg (default=“corr”) | Covariance calculation method: ordinary (Pearson product moment) covariance (“cov”), ordinary correlation (“corr”), Spearman rank covariance (“rcov”), Spearman rank correlation (“rcorr”), uncentered ordinary correlation (“ucorr”). Note that Kendall’s tau measures are not valid methods. |
pairwise | Compute using pairwise deletion of observations with missing cases (pairwise samples). |
df | Compute covariances with a degree-of-freedom correction accounting for the mean (for centered specifications). The default behavior in these cases is to perform no adjustment ( e.g. – compute sample covariance dividing by  rather than  ). |
Examples
ser1.panpcomp(eigval=v1, eigvec=m1)
computes the principal components decomposition of the contemporaneous (between cross-section) Pearson correlation matrix for the series SER1.
The output view is stored in a table named tab1, the eigenvalues in a vector named v1, and the eigenvectors in a matrix named m1.
ser1.panpcomp(out=graph)
ser2.panpcomp(out=graph, scree, cproport)
displays a screen plot of the eigenvalues, and a graph containing both a screen plot and a plot of the cumulative eigenvalue proportions.
ser.panpcomp(period, cov=rcorr, out=loading)
displays a loadings plot for the principal component decomposition of the period (within cross-section) Spearman rank correlation matrix, and
ser.panpcomp(period, cov=rcorr, out=biplot, scale=symmetric, mult=lt) 1 2 3
displays a symmetric biplot of the period Spearman correlation matrix for all three pairwise comparisons.
Cross-references
To compute principal components scores and save them in series in the workfile, see
Series::makepanpcomp.