questions about p value

[liuyu988]

Hi, Thanks for your tools! I don't make sense of the parameter --cstat. When I use different float to define '-cstat', I got different numbers of differential AS events, and one AS event maybe is differential , but it isn't in other case, but the Inclevel value don't change in this two cases.So what the real meaning --cstat is?

[EricKutschera]

--cstat is the cutoff used in the statistical test. It changes the PValue and FDR columns which are output by the statisical test

The IncLevel columns do not depend on the statistical test. The inclusion level is calculated from the read counts for each isoform and it doesn't depend on --cstat

See this similar post: https://github.com/Xinglab/rmats-turbo/issues/47#issuecomment-694843511

[liuyu988]

Thanks a lot for your rapid reply! so if I set --cstat equal to 0, i can detect not only events with very small IncLevelDifference, but also with bigger IncLevelDifference. I don't know if I understand this issue correctly, so any help would be appreciated.

Yu

[EricKutschera]

Regardless of the value for --cstat rMATS will produce the same set of events in the output files. If --cstat 0 is used then the PValue column is evaluating whether there is any difference in the inclusion level between the sample groups for each event. If --cstat 0.1 is used then the PValue column is evaluating whether there is at least a 0.1 difference in the inclusion level. A low value like --cstat 0 is an easier condition and will result in more significant PValues. A higher value like --cstat 0.1 is a more difficult condition and will result in less significant PValues.

Here is a thread discussing cstat: https://groups.google.com/g/rmats-user-group/c/RIGKPwXK9eI/m/KlAJACN4BAAJ