Preranked metric ties

[ToledoEM]

Hi, great package. I am wondering how it is manage when the metric of the ranked genes it is a tie, this for values not equal to 0.

The Broad Institute java software has a warning about this, but I did not find any information about it in the preprint nor the vignette.

Should ties be solve with jittering the duplicated values? or a warning be printed when there are ties in the ranked gene list?

[assaron]

Hi,

For now ranking ties are not properly supported and I guess the behavior is pretty similar to Broad's. While the results could be somewhat errorneous, I don't think it really matters in most circumstances. Moreover, there is no point in jittering, the current behaviour is equivalent to jittering ("he order of genes will be arbitrary").

I had a similar request recently and I think I know how to properly deal with it, but don't have time to implement it yet.

[ToledoEM]

Hi, You are right about jittering, same output. Maybe a warning should be enough.
Thanks

[assaron]

I'll keep the issue open until it's resolved properly.

[mayanbriller]

Hi! I am also having this problem. I think it originates from using something like p-values (for me) as a ranking metric in a pre-ranked gene list. In the original GSEA you can choose to do a weighted analysis (under enrichment statistic) or 'classic' - which is the right one to choose when using a metric like p-value (where the actual value doesn't matter I guess)