Partial spearman correlations and normality

#1
hello

I am trying to conduct partial spearman rho correlations with syntax. However, I have different controls for my different variables. For example, with one variable I have 3 controls and another variable I have 5 controls. I am using the syntax from here: http://www-01.ibm.com/support/docview.wss?uid=swg21474822 but can't seem to be able to run it when I have different controls. Is there a way I can do this or alter the syntax for me to run it like this?

Additionally, I wondered if anyone could advise on sample size and normality. I have a sample size of 220, with non normal distributions. Does anyone have references that would be helpful in suggesting that this sample size is big enough to overcome the assumptions of normality to run linear regression?

many thanks

Loren
 

Dragan

Super Moderator
#2
The Spearman correlation is really nothing more than applying the formula associated with Pearson correlation to the ranks of the data. A sample size of 220 would be appropriately large enough to invoke the central limit theorem. As such, if you want to use Spearman correlations just use the following expression:

\( r_{s}=\frac{6}{\pi }\arcsin \left ( \frac{1}{2}r_{p} \right ) \)

For a proof of this see here (at the very bottom), I'm just taking the inverse of the result: http://www.talkstats.com/showthread.php/53229-Proof-of-the-day