Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
NCTM or The Math Forum.



How to calculate variance explained by clusters (multivariate)?
Posted:
Nov 1, 2012 10:14 PM


I want to calculate the variance explained by clusters. For univariate data, it is easy: betweengroup variance / total variance.
But what about for multivariate data? Can I just use the same equations as for betwengroup SS and degree of freedom, but replace (xixj)^2 terms with d(xi,xj)^2 terms?
So for instance, in univariate the numerator for total variance would be: sum[(xixmean)^2] For the total variance of my multivariate data can I just use: sum[d(xixmean)^2] ?
Or is there more to it than that? Can anyone recommend a nice description of this that actually gives equations?



