[BioC] limma with missing values

Gordon K Smyth smyth at wehi.EDU.AU
Tue Mar 11 00:11:39 CET 2008



On Sat, 8 Mar 2008, Hans-Ulrich Klein wrote:

>> limma treats missing values in exactly the same way as other linear model 
>> function in R, such as lm(), glm() etc.  For each gene, the rows with 
>> missing values are removed both from the data and design matrix, then 
>> non-estimable coefficients are removed (given NA) from the end of the 
>> design matrix.

> The problem is that often a user will not notice when limma drops a column of 
> the design matrix for a few of many thousand genes. Maybe a warning message 
> should be printed?

I have now added a warning message to lmFit() in this situation.

Best wishes
Gordon



More information about the Bioconductor mailing list