RE: st: Spearman correlation with adjustment

["Tiago V. Pereira" <tiago.pereira@mbe.bio.br>]

Cecilia,

You might explore the following approach (and see if it makes some sense
in your case):

Assumption:  there are no ties. So, you can compute  spearman's
coefficient (rho_S) from  pearson's coefficient (rho_P)

Approach:

1) Create two or more categories or subgroups in which the confounding
variable has a smaller role
2) Within each category compute ranks for your values
2) Calculate the pearson coefficient using those ranks (that is, rho_P
will be calculated from ranked variables)
3) transform the rho_Ps into Z scores (r to z' transformation - Fisher
approach)
4) perform a meta-analysis of Z scores
5) get the results back to the original metric (rho_P)

This approach is likely to provide less biased results compared to raw
analyses. It also provides the opportunity to check/quantify if there is
statistical heterogeneity among subgroups (Cochran's Q test, I^2 index).

All you need is the rho_P from ranked variables and the following packages
-corrci- (or -corrcii-) and -metan-

also check:
http://mason.gmu.edu/~dwilsonb/ma.html
http://www.stata.com/statalist/archive/2010-06/msg00728.html

Cheers!

Tiago




*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/