search

stephenslab / mashr

An R package for multivariate adaptive shrinkage.
https://stephenslab.github.io/mashr
Other
87 stars 19 forks source link

Pairwise sharing - why does effect only have to be significant in one condition? #86

Open arider2 opened 3 years ago

arider2 commented 3 years ago

Hello,

This isn't so much an issue but a general question - why does an effect (i.e eQTL) only have to be significant (i.e. lfsr 0.05) in one condition for it to be classed as shared between the conditions? Is it too stringent to ask for an effect to be significant in both? Does this underestimate sharing?

Thank you.

stephens999 commented 3 years ago

sorry I don't understand the question. What do you mean by "classed as shared between the conditions"? Is there a particular function or part of the documentation you are referring to?

On Mon, Sep 28, 2020 at 11:23 AM arider2 notifications@github.com wrote:

Hello,

This isn't so much an issue but a general question - why does an effect (i.e eQTL) only have to be significant (i.e. lfsr 0.05) in one condition for it to be classed as shared between the conditions? Is it too stringent to ask for an effect to be significant in both? Does this underestimate sharing?

Thank you.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/stephenslab/mashr/issues/86, or unsubscribe https://github.com/notifications/unsubscribe-auth/AANXRRJ4WHVLO4K75DYBVGTSIC2A5ANCNFSM4R4WYN6Q .

arider2 commented 3 years ago

Sorry for the lack of clarity.

I'm referring to the get_pairwise_sharing function, which calculates the proportion of effects that are shared between each pair of conditions.

If I understand this function correctly, for each pair of conditions it first identifies the effects that are significant in at least one of the two conditions, where significance is defined by the user with the lfsr_thresh parameter.

My questions is why doesn't this function identify effects that meet the lfsr threshold in both conditions? I.e. why identify the union of significant effects instead of intersect?

Hope this makes sense.

sorry I don't understand the question. What do you mean by "classed as shared between the conditions"? Is there a particular function or part of the documentation you are referring to? … On Mon, Sep 28, 2020 at 11:23 AM arider2 @.***> wrote: Hello, This isn't so much an issue but a general question - why does an effect (i.e eQTL) only have to be significant (i.e. lfsr 0.05) in one condition for it to be classed as shared between the conditions? Is it too stringent to ask for an effect to be significant in both? Does this underestimate sharing? Thank you. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#86>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AANXRRJ4WHVLO4K75DYBVGTSIC2A5ANCNFSM4R4WYN6Q .

stephens999 commented 3 years ago

you are correct, it first identifies the effects that are significant in at least one of the two conditions, where significance is defined by the user with the lfsr_thresh parameter. Let's call this number N.

Then assesses which of these N effects are shared. It does that by either looking at shared sign (S) or shared magnitude (M).

Then the sharing is computed as S/N or M/N. So N is the denominator in the computation, not the numerator. It is asking what proportion of the eQTLs in at least one condition are shared.

If you computed N only as the ones significant in both conditions you would be effectively asking what proportion of the shared eQTLs are shared.... which would not make sense.

Does that help?

On Mon, Sep 28, 2020 at 11:50 AM arider2 notifications@github.com wrote:

Sorry for the lack of clarity.

I'm referring to the get_pairwise_sharing function, which calculates the proportion of effects that are shared between each pair of conditions.

If I understand this function correctly, for each pair of conditions it first identifies the effects that are significant in at least one of the two conditions, where significance is defined by the user with the lfsr_thresh parameter.

My questions is why doesn't this function identify effects that meet the lfsr threshold in both conditions? I.e. why identify the union of significant effects instead of intersect?

Hope this makes sense.

sorry I don't understand the question. What do you mean by "classed as shared between the conditions"? Is there a particular function or part of the documentation you are referring to? … <#m-8111842325393127155> On Mon, Sep 28, 2020 at 11:23 AM arider2 @.***> wrote: Hello, This isn't so much an issue but a general question - why does an effect (i.e eQTL) only have to be significant (i.e. lfsr 0.05) in one condition for it to be classed as shared between the conditions? Is it too stringent to ask for an effect to be significant in both? Does this underestimate sharing? Thank you. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#86 https://github.com/stephenslab/mashr/issues/86>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AANXRRJ4WHVLO4K75DYBVGTSIC2A5ANCNFSM4R4WYN6Q .

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/stephenslab/mashr/issues/86#issuecomment-700154790, or unsubscribe https://github.com/notifications/unsubscribe-auth/AANXRRPNJBKQCNEWDTAB2OTSIC5FXANCNFSM4R4WYN6Q .