@KirkegaardEmil @gwern I had often wondered about lasso, horseshoe priors. do you have a reference for this?
-
@KirkegaardEmil@_r_c_a Looks like they don't directly compare to OLS one-at-a-time estimates? Closest seems to be the rrBLUP algorithm.
-
-
@KirkegaardEmil@_r_c_a I was going to simulate some data for IQ and see if I could use JAGS or something. How hard could a GWAS be to code? -
@gwern@KirkegaardEmil realistic SNPs are the hard part, isn't it? Horseshoe prior in rstanarm or brms v easy to do. -
@_r_c_a@KirkegaardEmil Nah. You just fit an exponential distribution to Rietveld's betas; then you can easily generate sets of SNPs. -
@gwern@KirkegaardEmil what about allele freq and linkage diseq.. -
@_r_c_a@KirkegaardEmil Allele freq usually ~0.5 for top hits. LD I just ignore; causal tagging is more important, and is a mixture model. -
@gwern@KirkegaardEmil so you make it easy :-) maybe defensible, can't really assess. Is this standard practice? -
@_r_c_a@KirkegaardEmil I think defensible: several SNPs in LD means effect is split between, implying flatter dist & underestimate of gains -
@gwern@KirkegaardEmil http://biorxiv.org/content/early/2016/01/10/034082 … 0.8% gain for eg. t2d with lasso
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.