[BioC] Fitting a contrast matrix in Limma

[Gordon Smyth]

>Laura Hollins LHollins at PICR.man.ac.uk
>Wed May 26 17:52:29 CEST 2004
>
>Hi,
>
>I have been trying to fit a contrast matrix in Limma using the
>'contrasts.fit' function but I keep getting an error:
>
>Error in La.chol(crossprod(design)) : the leading minor of order 5 is
>not positive definite
>
>I entered my design and contrast matrices from tab-delimited files
>created on Excel using the 'read.delim' function and I then had to
>convert them into matrices from data frames. In the past, I have typed
>smaller, less complex design and contrast matrices directly into R and
>had no problems.
>
>Does anyone know what this error means or how to resolve it?

It means that your design matrix is not of full rank. This being the case, 
you should have got an error message from lmFit() or lm.series() when you 
created your fit object in the first place. In other words, you've done 
something wrong because you should never get to the stage of applying a 
contrasts matrix if your design matrix is singular.

Hard to give anymore help than this because you don't give your actual 
commands.

Gordon

>Thanks,
>
>Laura