cpm vs total normalization dramaticallly change results - why?

[Piotr Bragoszewski]

Hi,

Using PinAPL-Py to analyze my data, I have noticed that by changing only normalization between CPM and total, I get significantly different results with very different p-value distribution. In both cases, I have used aRRA and fdr_bh. I don't see a reason for this effect. 

My samples are not equal in reads numbers 11M, 11M, 15M, and 6M.

What might be the cause?

Best,

Piotr

[Philipp Spahn]

Hi Piotr,

changing the normalization can absolutely change the results, in particular when read depths vary between samples.
As far as the gene ranking method is concerned, I would consider trying the default method rather than aRRA because the former relies less on statistical assumptions and is more straightforward to interpret.

Kind regards,
Philipp

--
You received this message because you are subscribed to the Google Groups "PinAPL-Py" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pinapl-py+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/pinapl-py/920c66c0-02b7-494b-a2e8-5b99c2ada5e3n%40googlegroups.com.