[R] significance levels for partial correlations?

Ashish Ranpura buddhahead at ranpura.com
Tue Nov 20 12:48:05 CET 2007


I've seen that this question has been asked before, in the archives,  
but I haven't been able to find a workable answer. This may be a  
failure to understand the statistics!

The problem is that, while I can easily get partial correlation  
values out of corpcor, how do I get significance values? Surely the  
significance of a partial correlation between two variables isn't the  
same as the simple bivariate correlation significance (cor.test)? As  
I said, I may be misunderstanding the stats.


And just to round out the scenario, I should mention that I've  
written a function that uses the corpcor library to perform partial  
correlations on "experimental" variables while partialing out the  
effect of "control" variables (as opposed to getting the partial  
correlation of a single variable controlling for ALL other  
variables). This is just a convenience function that allows me to  
stick in a whole data frame and analyze only those variables I'm  
interested in.

I won't paste in that function here, it's not relevant to the  
problem, but if anyone wants a copy just let me know. Here is the  
syntax:

pcor.n = function (df, cont, expi="") {
###
## Perform an n-order partial correlation, controlling for the  
influence of
##   control variables on experimental ones.
##
## df		data frame containing all the variables.
## cont		a list of control variable names (list of strings)
## expi		a list of experimental variable names (list of strings).
##			   (defaults to all variables excluding controls)
##
## RETURN VALUES
##   The function returns a matrix of partial correlation values,  
excluding the
##   effects of the control variables.
##
###

Thanks for any guidance,

-Ashish Ranpura.


-----
Ashish Ranpura
Institute of Cognitive Neuroscience
University College London
17 Queen Square
London WC1N 3AR

tel: +44 (20) 7679 1126
web: http://www.icn.ucl.ac.uk



More information about the R-help mailing list