Given a matrix of phi or Yule correlation coefficients and a vector of marginals, use the polychoric function to convert these to polychoric correlations. Some older correlation matrices were reported as matrices of Phi or of Yule correlations. That is, correlations were found from the two by two table of counts:
lll{
a b
c d
}
Yule Q is (ad - bc)/(ad+bc).
With marginal frequencies of a+b, c+d, a+c, b+d.
Given a square matrix of such correlations, and the proportions for each variable that are in the a + b cells, it is possible to reconvert each correlation into a two by two table and then estimate the corresponding polychoric correlation (using John Fox's polychor function.