Correlation matrix for MDS (phi-squared)

[Sophie Suda]

Dear all,

for my multi-dimensional scaling of 10 variables I want to use a phi-square measure for the (dis-)similarity matrix. Taking other measurements as an alternative is not really an option because I am replicating someone else's work who did their calculations in SPSS.

For the basic mds command
mds var1-var10, id(id) method(classical)
there is the option measure(x) to choose the matrix option, but phi-square is not available.


I also couldn't find any package as an alternative.
Is there a way to build the correlation matrix first and then do the mds step-by-step (which probably shouldn't be a problem with mdsmat after getting the matrix)? Or does anyone know about any packages that do the job for me?


Thank you all in advance!

[George Ford]

Maybe this would do it.

net describe snp3, from(http://www.stata.com/stb/stb3)

Perhaps the

method(Pearson)

would get you phi and then you can square it? Not sure if it's the same thing.

G