Correlation: is this correct?

ans ans ans

Answer 1

[George Papazafeiropoulos]

Dear Nuchto,

See this:

x=rand(4,30); y=10*x+5; corr=sum(zscore(x',1).*zscore(y',1))'/size(x,2)

this is different from your version, where zscore scales X using the sample standard deviation, with n - 1 in the denominator of the standard deviation formula. zscore(X,0) is the same as zscore(X). In the above version zscore(X,1) scales X using the population standard deviation,with n in the denominator of standard deviation formula. This is the right way since the correlation coefficient r is given by [Σ(xi*yi)-(Σxi)*(Σyi)/N]/{[Σxi^2-(Σxi)^2/N]*[Σyi^2-(Σyi)^2/N]}^(1/2) and after some mathematical operations gives the above formula.

Greetings,

George Papazafeiropoulos

[Nuchto]

Thanks! I actually never understood where this size(x,2) comes from. I thought it should be already included in the zscore part. But my knowledge of this is very basic.

[George Papazafeiropoulos]

Alternatively, look at the function corrcoef(X). Good luck!

George Papazafeiropoulos

[George Papazafeiropoulos]

Because in the denomimators of the original formula N appears, not N-1.