search

stephenslab / susieR

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

default for purity #57

Closed stephens999 closed 5 years ago

stephens999 commented 5 years ago

I think we should use r^2 0.8 as the default

first, that is a more usual default for "high" correlation in association analysis

second i think it is more appropriate for general use. Any default will be subjective, but the current default is very low IMO.

stephens999 commented 5 years ago

Further thought on this, I think my suggestion was misguided. It is better to have a lower threshold, so we don't remove something that is of potential interest, and let the user decide having been presented the results.

However I don't think we should justify the current choice as a widely used threshold

gaow commented 5 years ago

Oh great! Yes indeed default threshold is subjective, but I also believed that lower threshold is better. Was reluctant to change to r^2 0.8 (r 0.9) because it feels a bit too high.

However I don't think we should justify the current choice as a widely used threshold

what would you suggest then (and why)? DAP-G's suggestion is r^2 0.25 but maybe you do not think it is a good "widely used" (outside context of fine-mapping) threshold either.

stephens999 commented 5 years ago

i'd suggest we leave default threshold as is for now and just document what the default threshold is. no need to say it is widely used.

On Fri, Oct 19, 2018 at 8:11 AM gaow notifications@github.com wrote:

Oh great! Yes indeed default threshold is subjective, but I also believed that lower threshold is better. Was reluctant to change to r^2 0.8 (r 0.9) because it feels a bit too high.

However I don't think we should justify the current choice as a widely used threshold

what would you suggest then (and why)? DAP-G's suggestion is r^2 0.25 but maybe you do not think it is a good "widely used" (outside context of fine-mapping) threshold either.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/stephenslab/susieR/issues/57#issuecomment-431358194, or mute the thread https://github.com/notifications/unsubscribe-auth/ABt4xcE37J_XsIABZWa_yU7p4Dh83l-vks5umc-VgaJpZM4XulrV .

stephens999 commented 5 years ago

fixed by changing documentation