R help - Dec 2008 - Using a covariance matrix as input to relaimpo package

I'm having trouble getting the relaimpo package to use a covariance matrix
as
input.  I'm getting an error message that reads as follows:

Error in eval(m$weights, data, parent.frame()) : 
  numeric 'envir' arg not of length one

I'm guessing there is something wrong with the structure of my covariance
matrix, but it looks fine to me.  Pardon my R ignorance if this is an easy
error message to decipher.  Below is my code.

# calculate covariance matrix from survey respondent data using pairwise
deletion
covmatrx
cov(respdata[,c("V0007","V0029","V0031","V0032","V0034","V0035","V0036")],
use = "pairwise")
> 
# show the resulting covariance matrix
> covmatrx
           V0007       V0029       V0031        V0032       V0034      
V0035       V0036
V0007 0.5713399 0.4207775 0.4144349 0.4178506 0.3679451 0.4073394 0.3847520
V0029 0.4207775 0.7043533 0.5567348 0.5877138 0.5052262 0.5627425 0.5289048
V0031 0.4144349 0.5567348 0.6335806 0.5735619 0.4971775 0.5453335 0.5179622
V0032 0.4178506 0.5877138 0.5735619 0.6922080 0.5206816 0.5638391 0.5579144
V0034 0.3679451 0.5052262 0.4971775 0.5206816 0.5369755 0.4925769 0.4754083
V0035 0.4073394 0.5627425 0.5453335 0.5638391 0.4925769 0.5977329 0.5098200
V0036 0.3847520 0.5289048 0.5179622 0.5579144 0.4754083 0.5098200 0.6480419

# try the lmg method of relative importance
> imps1 = calc.relimp(V0007 ~ V0029 + V0031 + V0032 + V0034 + V0035 + V0036,
> data=covmatrx, type="lmg", rela=TRUE)
Error in eval(m$weights, data, parent.frame()) : 
  numeric 'envir' arg not of length one
> 
-- 
View this message in context:
http://www.nabble.com/Using-a-covariance-matrix-as-input-to-relaimpo-package-tp21022295p21022295.html
Sent from the R help mailing list archive at Nabble.com.

I just discovered that even with this error message, relaimpo does run and
produce output objects that appear fine.  It would be nice to know if this
error message is serious or not.

-- 
View this message in context:
http://www.nabble.com/Using-a-covariance-matrix-as-input-to-relaimpo-package-tp21022295p21022374.html
Sent from the R help mailing list archive at Nabble.com.

By trial and error I have discovered that it works if I don't use the
formula
interface in combination with a covariance matrix as input.

If the covariance matrix has the dependent variable as its left-most
variable as the relaimpo documentation suggests, then the relaimpo package
will run by simply naming the covariance matrix as the first object in the
call and not using a formula.  The downside of this is needing to create
different covariance matrices for different models.

The following will work:

# calculate covariance matrix from survey respondent data using pairwise
deletion
covmatrx
cov(respdata[,c("V0007","V0029","V0031","V0032","V0034","V0035","V0036")],
use = "pairwise")

# try the lmg method of relative importance
> imps1 = calc.relimp(covmatrx, type="lmg", rela=TRUE)

-- 
View this message in context:
http://www.nabble.com/Using-a-covariance-matrix-as-input-to-relaimpo-package-tp21022295p21041633.html
Sent from the R help mailing list archive at Nabble.com.