R devel - May 2008 - Using option na.rm=True in function SD does not work for matrix with complete columns of NAs (PR#11364)

Dear R-developers,
=20
according to the "what's new"-section in version R 2.7.0, there
has been a change in the working of co[rv] and so also in sd and var.
=20
I am afraid, that the use of function sd with option "na.rm=3DT" has
not been changed appropriately. So the following problem exists with missing
data:
=20
> sessionInfo()
R version 2.7.0 (2008-04-22)=20
i386-pc-mingw32=20
=20
locale:
LC_COLLATE=3DGerman_Austria.1252;LC_CTYPE=3DGerman_Austria.1252;LC_MONETARY=3DGerman_Austria.1252;LC_NUMERIC=3DC;LC_TIME=3DGerman_Austria.1252
=20
attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base  =20
=20
=20
a<-matrix(rnorm(100),10,10); a[1,]<-NA
> sd(a,na.rm=3DT)
Fehler in var(x, na.rm =3D na.rm) : keine vollst=E4ndigen Elementpaare

this error is clear as "sd" is using "var" and in case of
na.rm=3DT "var" uses "na.method=3Dcomplete.obs" which is
obviously never the case in this example.
=20
So I suggest to set the default.method for na.rm=3DT to
"pairwise.complete.obs" to get the same result as in previous versions
of R, that is the same as=20
> sqrt(diag(var(a,use=3D"pair")))
 [1]        NA 0.9117738 0.5953496 0.9272141 0.8864333 0.8045383 0.4240055
1.0304519 0.9221493 1.2087638

Best regards,
=20
Martin

--=20
Martin Tschiedel
Senior Investment Manager	 A Siemens Company=09

INNOVEST Kapitalanlage AG
Email: m.tschiedel at innovest.at