Closed joshchiou closed 3 years ago
Are you using susie or susie_rss?
On Wed, Jun 2, 2021, 00:29 Josh Chiou @.***> wrote:
Is there a way to recover (or approximate) the per-variant effect size estimates for each signal/credible set? From my limited understanding, mu contains the posterior z-scores for each signal. Is there a simple solution that I'm missing here? (e.g. mu * sqrt(posterior variance))?
Thanks, Josh
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/stephenslab/susieR/issues/129, or unsubscribe https://github.com/notifications/unsubscribe-auth/AANXRRNH5YCFZG5WP6ZFPU3TQW6TNANCNFSM456AOISQ .
I'm using susie_rss.
To get effect estimates use 'susie_get_posterior_mean'
If you used susie_rss you need to multiply those estimates by the original standard errors (ie by s_j if z_j= bhat_j/s_j) to get estimates on the original scale.
Note that these are unconditional effect estimates (ie scaled by the pips). Thus if a snp is in ld with a lot of other snps it's unconditional effect estimate will tend to be small because its pip will tend to be small.
On Wed, Jun 2, 2021, 18:24 Josh Chiou @.***> wrote:
I'm using susie_rss.
Are you using susie or susierss? … <#m-3668851743586760740_m_8177645735469955484_m8895739208112277259> On Wed, Jun 2, 2021, 00:29 Josh Chiou @.**> wrote: Is there a way to recover (or approximate) the per-variant effect size estimates for each signal/credible set? From my limited understanding, mu contains the posterior z-scores for each signal. Is there a simple solution that I'm missing here? (e.g. mu sqrt(posterior variance))? Thanks, Josh — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#129 https://github.com/stephenslab/susieR/issues/129>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AANXRRNH5YCFZG5WP6ZFPU3TQW6TNANCNFSM456AOISQ .
I'm using susie_rss.
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/stephenslab/susieR/issues/129#issuecomment-853445869, or unsubscribe https://github.com/notifications/unsubscribe-auth/AANXRRLZ7TSJK2MVJEXGODLTQ24UFANCNFSM456AOISQ .
Is there a way to recover (or approximate) the per-variant effect size estimates for each signal/credible set? From my limited understanding, mu contains the posterior z-scores for each signal. Is there a simple solution that I'm missing here? (e.g. mu * sqrt(posterior variance))?
Thanks, Josh