Analysis(statistical)??

Question

-though "t-test of correlation(bivariate)seems the likely choice,is it(for this sort of prob:\):
"..finding significance of "inter(??)-correlation" between education level and marital satisfaction."
Sounded like simple correlation to me..??THANKS in advance~

Answer 1

If I understand the question, you want to regress the variable y = "marital satisfaction" on x = "education level", compute the sample correlation coefficient r, and then test the null hypothesis, H0, that the true correlation coefficient, rho, is zero (or maybe you have some other value in mind). If H0 is that rho = 0, you can use the statistic

t = r sqrt(n-2) / [sqrt(1 - r^2)]

where r is the sample correlation coefficient (and n is the sample size). Under H0, this t has Student's distribution with n-2 degrees of freedom. This fact can be used for a test in the usual way.

But if H0 is that rho = rho_0, where rho_0 is some specific *nonzero* value, then the above t is *not* Student-distributed and you'll need to use Fisher's Z.
The point of Fisher's transform Z(r) is that it is approximately normally distributed, with mean Z(rho) and std dev. 1/sqrt(n-3), no matter what rho is. (So one can use it to test "rho = rho_0" with a *nonzero* rho_0, or to give a confidence interval for rho. Of course, it can also be used to test "rho = 0", but for this the above t will do and is simpler.)
For how to use Fisher's Z, see page 20 of http://www.statpower.net/Content/310/Unified%20Approach.pdf

Answer 2

You might want to try Pearson's method of testing correlation significance. Check out the web site I've listed as my source. It's got a lot of statistical methods and advice.