search

VEuPathDB / microbiomeComputations

1 stars 0 forks source link

Correlation returning NaN?? #73

Closed asizemore closed 8 months ago

asizemore commented 9 months ago

Found this while working on another issue

Screen Shot 2023-12-19 at 6 34 24 AM

A big problem is that we're still rendering those edges! Possibly we're doing the number-string transformation wrong somewhere? Will have to check...

d-callan commented 8 months ago

is the correlation app currently displaying a count of how many records it represents? im wondering if the sample size here is small?

EDIT: I think i remember somewhere else saying wed do this on a per-edge basis only maybe?? if we remove the edges, thatd be very hard to do.

asizemore commented 8 months ago

No it's not displaying a count of how many records it represents per edge. Once we have the helper scatter we'll have that info! What if we set a min number of records each edge must use? Like a minimum "n". The computation can send that info to the frontend and the viz can add that info to the plot title. It could even be another filter if we wanted.

d-callan commented 8 months ago

Somewhere I think I made a ticket about prefiltering variables, and if this idea isn't in that ticket yet I think it should be.

d-callan commented 8 months ago

rare taxa where only a small number of samples have non-zero values are producing NaNs. to me, that is a valid result. id be in favor of letting the backend be, and improving handling frontend.

d-callan commented 8 months ago

also, for posterity, the issue about pre-filtering features is here https://github.com/VEuPathDB/microbiomeComputations/issues/69