[BioC] A question with DEXSeq package: inconsistency between normalized counts vs. fitted expression, fitted splicing or fold changes

[Alejandro Reyes]

Dear Yao Hui,

It was an error in the code, the levels in one of your variables in the 
design matrix were not sorted! At some point part of the code was 
(wrongly) relying on the order of the factors, some the conditions were 
mixed. You could stay with the same version and do this before starting 
the analysis:

levels( pData(testgene)$type ) <- sort( levels( pData(testgene)$type ) )

Or update to version DEXSeq_1.3.3? ( it is already fixed here).
An apology for the bug!

Alejandro