I was wondering if anyone was aware of any issue with running RUVfit (as part of the missMethyl package) when I have a large number of samples from my methylation experiment? I am working with a data set with 634 samples and 18 covariates, and I am getting the following error in the invvar function that is called by RUVinv: "In function invvar: The fourth-smallest eigenvalue is more than 10 times larger than the smallest eigenvalue. This is not currently supported, and very likely means something is wrong (perhaps a dimension was removed during preprocessing?)."
I believe this is happening because three of the eigenvalues computed in the invvar function are 0, which I think is the result of the number of control probes (613) being less than the number of independent components of the design matrix (634-18=616). Has anyone come across this before? Do you have any suggestions for a work-around? If I remove some samples, I no longer get the error, but I would of course rather increase the number of control probes instead. I just want to do so in a way that makes sense and won't break the functionality of the method, if possible.
Thanks for any help you can give me, it is greatly appreciated!