"Glen" <[EMAIL PROTECTED]> wrote in message
[EMAIL PROTECTED]">news:[EMAIL PROTECTED]...
> "Chia C Chong" <[EMAIL PROTECTED]> wrote in message
news:<a2fgce$b32$[EMAIL PROTECTED]>...
> > Hi!
> >
> > I have 2 random varaibles (X and Y) obtained from some experiments. I
have
> > expressed these 2 RVs in ternm of 2-D joint PDF f(X,Y)=f(Y|X)f(X).I
would
> > like to test the correlation between them to see whether there are
> > correlated or not. Do I simply find the correlation coeffient between
these
> > two variables or are there other ways that I could use to test
correlation??
>
> If you're interested in association rather than just correlation, take
> a look at f(Y|X) (*including* the range of values for y). If it
> doesn't depend on x, then the variables won't be associated.
> (Actually, the correct way to do it is to check that f(Y|X) = f(Y), so
> if you can find f(Y) easily, do that.)
Sorry if I being stupid here..what is the different between association &
correlation in their statistical term??
>
> Assessing if there's any linear correlation is a matter of checking if
> E(XY) = E(X) E(Y).
I have tried to find their correlation coefficient and its value is very
small.Howevevr, correlation coefficient only tells how linearly there are
correlated between these two values. Are there any kind of test for
correlation but neccessary linear correlation??
>
> What's the difference? As an example, imagine X is symmetric about 0,
> and let's further assume that at least the first few moments exist.
> Let Y=X^2.
>
> Then X and Y are perfectly associated (you tell me X, I'll tell you
> Y), but uncorrelated.
>
> Glen
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at
http://jse.stat.ncsu.edu/
=================================================================
