CHARMM Development Project Forums User Discussion & Questions General Chemistry Discussions depents test for multiple continous variables.

I have N sets of data with M variables. The variables are continous. I want to know if variable a is independent of the M-1 other variables. What stat test do I use? The chi-square test is sort of what I want but it is for non-continous, catagorized variables.

Assuming N > M and some observable property possibly related to the M variables, principal component analysis is one method which can be used.

My knowledge of PCA is rather poor but doesn't it assume linearity of some sort?

Am I right in thinking that if we have three variables, A, B, and C, that A can be independent of B and A can be independent of C individually but A can still be dependent on B and C together? (hope that makes sense.)

Heres a better why of saying it. If we know P(A|B)=P(A) and P(A|C)=P(A), can we say anything about P(A|B,C)?

Strictly speaking, I believe so, but it can give useful results regardless, and there are some non-linear PCA methods.

Perhaps cross correlation of the variable might be useful.

P(A|B,C)=P(A)