Dear all,
I have come across a very confusing matter regarding dist() supplied by
the amap package:
--- m is just a test matrix
> library(amap)
Loading required package: mva
Warning message:
package 'mva' has been merged into 'stats'
> m a b c
aa 0.1 0.2 0.3
bb 2.0 3.0 4.0
cc 2.0 4.0 6.0
dd 0.3 0.2 0.1
> ds<-dist(m,method="pearson")
> ds
aa bb cc
bb 7.416666e-03 cc 1.110223e-16 0.007416666
dd 2.857143e-01 0.205933333 0.2857143
> ds<-dist(m,method="correlation")
> ds
aa bb cc
bb -4.440892e-16 cc 6.661338e-16 0 dd 2.000000e+00 2 2
>
Why using method "correlation" I don't get 1-person also person
corr
coef is between (-1,1). Also in help it is said
>>'pearson': Also named "not centered Pearson" sum(x_i
y_i) /[sum(x_i2)
sum(y_i2)].
>>'correlation': Also named "Centered Pearson" 1 -
corr(x,y).
Maybe corr() is not Pearson correlation ?? Or maybe the data has to be
scaled? Very confusing. Maybe somebody can help me with this. Also there
is a sqrt missing in the person corr coef definition.
Best,
Radu
--
Dr. Radu Dobrin
Department of Molecular Biology
Princeton University
Washington Road
Princeton, NJ 08544-1014
Phone: 609-258-5657
E-mail: rdobrin at molbio.princeton.edu
