Canonical Correlation (Package CCA)

#1
Hi there,
I'm using R (and package CCA) trying to perform a regularized canonical correlation analysis (There are 12-13 "response" variables in one dataset, another dataset of a whole lot (i.e. >400) of potential "explanatory" variables, but only a sample size of 15). Gonzalez et al. (2008, http://www.jstatsoft.org/v23/i12/paper) mention that the package "includes a regularized version of CCA to deal with data sets with more variables than units", which is certainly what I have with only 15 "units" and I have been following the process they go through in the paper. But I get to an error message, and I do not know what to do about it. Here is the code, and any ideas or knowledge as to the subject would be appreciated.
Response variables are in matrix Y and explanatory variables are in matrix X

library(CCA)
correl <- matcor(X, Y)
img.matcor(correl, type = 2)
res.regul <- estim.regul(X, Y, plt = TRUE,
grid1 = seq(0.0001, 0.2, l=51),
grid2 = seq(0, 0.2, l=51))

(Note: estim.regul() takes a long time (~30-40 min) to complete when you use the sample data nutrimouse from CCA).

With my data is used in the estim.regul() step, I get the following:
Error in chol.default(Bmat) : the leading minor of order 12 is not positive definite

Does anyone know what to do about this error? Is it because some of my columns have an NA in them, or something else? Could it be due to columns with too many 0's?