Once bias is introduced, we enter the dangerous realm of self-fulfilling prophecies.
Vicious circle
Winner's curse.
Prior knowledge in the form of pathway databases. For example we cannot distinguish between missing data and false negatives. If something is not there does it mean it has not been found yet or does it truly not exist?
Posterior feedback into priors
In GWAS associated SNPs are preferentially annotated (Mike Keale). Does the annotation add any value?
Minor allele frequencies obtained from previous studies guide the clustering of genotypes in current studies. Will this online updating of probabilities eventually converge to the true MAF or are we reinforcing wrong freq estimates by influencing the influence of new data?
The posterior should become the new prior.
Double counting
No comments:
Post a Comment