Anders,
Does your data have missing values? It looks like they might. Look
at the 'use' parameter in cor. Also, is there a reason to use cor.test
instead of cor. Finally, if the expression values are not normal,
could you transform them first to make them more so--log2, for example?
And, no, no one has figured out the "best" way to define
distances/correlations for microarray, at least to the best of my
knowledge.
Sean
On Oct 14, 2004, at 8:37 AM, Anders Stegmann wrote:
> Hi, R!
>
>
>
> Question1:
>
> I am trying to correlate two vectors of numbers (two columns of
> microarray
> signal values) by using the non-parametric Spearman's rank correlation
> coefficient rho:
>
>> cor.test(V2.Signal,V3.Signal,method="spearman")
>
> but I get the error message:
>
> Error in if (q > (n^3 - n)/6) pspearman(q - 1, n, lower.tail = FALSE)
> else
> pspearman(q, :
> missing value where logical needed
> In addition: Warning message:
> NAs introduced by coercion
>
> I have tried to use the parametric Pearson correlation and the
> non-parametric Kendall's tau correlation and had no problem with that!!
>
>> cor(V2.Signal,V3.Signal,use="complete.obs")
>> (the
> Pearson correlation)
>
>> cor.test(V2.Signal,V3.Signal,method="kendall")
>> (the
> Kendall's correlation)
>
>
> what's wrong?
>
>
>
> Question2:
>
> Does anyone accidently know which correlation method would be the most
> correct to use when the microarray signal values (the values to be
> correlated) are not normal distributed (the Kendall's method seem to
> fit
> better to my other tests than the Pearson method).
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide!
> http://www.R-project.org/posting-guide.html