Entering edit mode
Can you email me a 'toy' example with your data of the failure?
On May 10, 2013, at 2:01 PM, Marc-Danie Jessica Nazaire wrote:
No, I am not.
Marc-Danie
On Friday, May 10, 2013 1:57:50 PM UTC-4, Evan Johnson wrote:
Are you using the 'sva' package version or the R script?
On May 10, 2013, at 1:51 PM, Marc-Danie Jessica Nazaire wrote:
Hi Igor,
We are getting this error when running ComBat with no covariates
specified (runs successfully with covariates) but there are no non-
numeric or missing values in our dataset.
I did some debugging and tracked down the error to this piece of code:
if (!NAs){gamma.hat <- solve(t(batch.design)%*%batch.design)%*%t(batch
.design)%*%t(as.matrix(s.data))}else{gamma.hat=apply(s.data,1,Beta.NA,
batch.design)}
NOTE: s.data and batch.design have no NAs
The gamma.hat matrix has NAs after this calculation and fails later
during the call:
gamma.bar <- apply(gamma.hat, 1, mean)
Do you have any more ideas or insights on what could be causing this
error?
Thanks,
Marc-Danie
On Tuesday, June 9, 2009 2:57:40 PM UTC-4, Evan wrote:
If you see the "Error in FUN(newX[, i], ...) : missing observations in
cov/cor" error message while running ComBat, search for non-numeric
or
missing values in your data file. It will work fine when you fixed
them.
Igor
--
[[alternative HTML version deleted]]
