Statistical help.....

[l***@yahoo.de]

Hello everyone!

I=B4m just working on my doc dissertation and I have a serious
statistical problem, maybe I can get help. Here=B4s my prob:

I=B4m just about to calculate Pearson correlations. I use a special type
of correlations because I got couple data, so I=B4m calculating z-Scores
as test sizes for every single correlation.

Because I=B4m expecting some correlations to be positive and some to be
negative, I =B4m mostly using one-sided statistical tests. What happens
when I=B4m expecting a positive correlation coefficient and my
calculated score is negative?

For example: I=B4m expecting a positive correlation, and the resulting
z-score is -1,15; does that mean that the p-Value is about 0.87
(because it=B4s in the opposite direction of my hypothesis?). In many
statistical books it=B4s said that in the tables describing the standard
normal distribution it=B4s just reversed for negative values, but that
can=B4t be true for one-sided tests because in that case the p-Value
would be about 0.13......

Please help me with that, I=B4m quite confused now.=20

Thank you

Lars



'`'`'`'`'`'`'`'`'`'`'`'`'`''`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`
sci.psychology.research is a moderated newsgroup.
Before submitting an article, please read the guidelines which are posted
here bimonthly or the charter on the web at http://psychcentral.com/spr/
Submissions are acknowledged automatically.

[s***@yahoo.com.au]

A serious statistical problem? Is that in the DSM? ;-) Doesn't seem
too serious really ...

Yes, a score of -1.15 would imply a p of 0.8749 (using the normal
distribution). The hypothesis you want to put to the test is that the
correlation is +ve. This means your null hypothesis is that r<=0,
while the alternative hypothesis is that r>0. The p value you gave as
an example indicates (in theory at least) that it is more likely the
null hypothesis is correct -- specifically, there's about a 0.87 chance
the null hypothesis is correct but only a 0.13 chance the alternate is
correct, given the observed r and provided it is appropriate to use the
normal distribution.

(note that if the z-score was 1.15, there would be 0.13 chance that the
null is correct, which is where the reversing comes in ... and if there
is less than a 0.05 chance that r<=0, you'd 'reject' the null in favor
of the alternate hypothesis that r>0 according to the standard view of
these things)

Do you have a large sample? For Pearson correlations, the student's t
distribution is usually used, but of course the t distribution tends
toward the normal for large N. Or are you using a z-score because of
the type of correlation?



'`'`'`'`'`'`'`'`'`'`'`'`'`''`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`'`
sci.psychology.research is a moderated newsgroup.
Before submitting an article, please read the guidelines which are posted
here bimonthly or the charter on the web at http://psychcentral.com/spr/
Submissions are acknowledged automatically.