[R] correlation with ties & multiple comparisons problem

Joseph Kambeitz jkamb at gmx.de
Thu Mar 19 11:00:29 CET 2009


I am currently working on a big data set. There are different subjects 
that have been measured in about 40 different variables (in this case a 
measure of brain metabolism). Now I want to look at possbile significant 
correlations between those variables. As some of the variables are not 
normally distributed I used "spearman" as method in the function 
rcor.test from the ltm-package. Unfourtunatly this results in more than 
50 warnings like this due to ties:

In cor.test.default(mat[, index[i, 1]], mat[, index[i,  ... :
  Cannot compute exact p-value with ties

I have now two question:
1. Is there any way to calculate a exact p-value for data that is not 
noramlly distributed?
2. Do have to correct for multiple comparisons in this case as I am 
practically testing each possible pair of the 40 variables?

Many thanks for your help!

More information about the R-help mailing list