runsusie() error with convergence
jvalliere opened this issue · 1 comments
jvalliere commented
Hello, I was wondering why in the runsusie() function convergence must be necessary in order to continue forward with colocalization. I had been running susie with two phenotypes in a region with ~6000 variants, and one was not reaching convergence. When I run susie_rss() myself with the data, the call does not converge but I get credible sets, which I thought could still be used for coloc. I then used annotate_susie() to make it compatible for coloc, and it worked fine. Is there a reason coloc must ensure convergence? Thank you so much for your help!
chr1swallace commented
typically if we fit an MCMC model, we require convergence so that we can be sure we are sampling from the true posterior. if a model has not converged, we might be using nonsense in the downstream coloc. if you think it's good enough, so be it, but I would want to justify that decision somehow.
…________________________________
From: Jesse Valliere ***@***.***>
Sent: Wednesday, May 8, 2024 4:45 PM
To: chr1swallace/coloc ***@***.***>
Cc: Subscribed ***@***.***>
Subject: [chr1swallace/coloc] runsusie() error with convergence (Issue #163)
Hello, I was wondering why in the runsusie() function convergence must be necessary in order to continue forward with colocalization. I had been running susie with two phenotypes in a region with ~6000 variants, and one was not reaching convergence. When I run susie_rss() myself with the data, the call does not converge but I get credible sets, which I thought could still be used for coloc. I then used annotate_susie() to make it compatible for coloc, and it worked fine. Is there a reason coloc must ensure convergence? Thank you so much for your help!
—
Reply to this email directly, view it on GitHub<#163>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AAQWR2AUFX3Z7HDTEXNVCTDZBJCB3AVCNFSM6AAAAABHNHGIJ6VHI2DSMVQWIX3LMV43ASLTON2WKOZSGI4DKOJRGA4TMNI>.
You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>