NaN on fitNbinomGLMs using DESeq

[David [R]]

Hi,

I am doing some RNASeq analysis and, for some of the fitNbinomGLMs I get NaN scores on some of the factors
Any idea on why this would be and how to sort it out?
For info the design is kind of tricky: multi-fator with no replicas.
Thanks in advance forf your help on this

David

[Simon Anders]

You would need to show us your design and code, so that one can say anything

[David [R]]

code

Hi Simon,

Thanks for your post.
Please find here below the links towards the code and the RData itself.
I guess the problem comes from working on multi-factorial with no replicates, but I am not certain about it, nor about the best way to sort it out.
Thanks a lot.

David Rengel

The code:

https://echange-fichiers.inra.fr/get?k=qls8uKlzvbQdTvueLQE

The RData:

https://echange-fichiers.inra.fr/get?k=HfISmyxskNcBh62i6zG

[David [R]]

NA and not NaN

Hi,

Just to correct the previous post: it is NA that I get on my fits, and not NaN.
On the other hand, everytime I get them, they concern the "symptoms" factor. However, I do not get them everytime that facotr is in the model

Thanks!

David Rengel