plot_mds {corx} | R Documentation |
plot_mds
Description
Perform multidimensional scaling of a corx object and plot results
Usage
plot_mds(corx, k = NULL, abs = TRUE, ...)
Arguments
corx |
corx object |
k |
numeric. The number of clusters. If set to "auto" will be equal to the number of principal components that explain more than 5% of total variance. |
abs |
logical. If TRUE (the default) negative correlations will be turned positive. This means items with high negative correlations will be treated as highly similar. |
... |
additional arguments passed to ggpubr::ggscatter |
Details
plot_mds performs classic multidimensional scaling on a correlation matrix. The correlation matrix is first converted to a distance matrix using psych::cor2dist. This function employs the following formula:
d = \sqrt(2*(1-r))
These distances are then passed to stats::cmdscale where k = 2. To compute latex
, distances are predict from the cmdscale output and correlated with input distances. This correlation is squared.
If the value of R^2
is less than 70%, a warning will inform users that two-dimensions may not be sufficient to represent item relationships.
The position of variables is then plotted with ggplot2. Clusters of items are identified using stats::kmeans. The number of clusters is determined using principal component analysis unless specified.
References
Carlson, D.L., 2017. Quantitative methods in archaeology using R. Cambridge University Press.