search

stephenslab / susieR

R package for "sum of single effects" regression.
https://stephenslab.github.io/susieR
Other
176 stars 45 forks source link

coverage in susie_get_CS #69

Closed stephens999 closed 3 years ago

stephens999 commented 5 years ago

maybe we should return the claimed coverage (which may be > threshold) rather than the requested coverage?

gaow commented 5 years ago

@stephens999 what do you mean by this? Currently we simply copy around the input coverage. Are you suggesting we compute the sum of PIP in each CS and report a L vector of overages? Still it would be good to keep track of the requested coverage I guess. So maybe we can use a different variable name for the new information?

stephens999 commented 5 years ago

yes, that was my suggestion.

Do you think the requested coverage is important to keep?

On Tue, Oct 30, 2018 at 7:26 PM gaow notifications@github.com wrote:

@stephens999 https://github.com/stephens999 what do you mean by this? Currently we simply copy around the input coverage. Are you suggesting we compute the sum of PIP in each CS and report a L vector of overages? Still it would be good to keep track of the requested coverage I guess. So maybe we can use a different variable name for the new information?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/stephenslab/susieR/issues/69#issuecomment-434518559, or mute the thread https://github.com/notifications/unsubscribe-auth/ABt4xVvAX4uaSD4TWhKY1DXT6Ym5poMVks5uqO4rgaJpZM4YDHgL .

gaow commented 5 years ago

Well, it might not harm to keep it. A possible scenario is that sometimes we request a high coverage that can only be satisfied at the cost of purity -- that is we have a large CS with low purity. If the purity is too low it will get filtered out, leaving empty CS output (in other words SuSiE finds nothing). It would be less confusing to report requested coverage along with an empty results, compared to otherwise.