To explore¶

A note to particular issues I'd like to have vignettes for to clarify.

%revisions -s

To illustrate susie itself¶

[] How susie behaves when the top SNP is not the causal signal.
[] How susie copes with subtle difference in tightly correlated variants.
- with SER is perhaps enough, and provide some back-of-envelope BF computations.
- Look in N = 1 simulations of high LD > 0.95, PIP difference > 0.8. This may also indicate small difference in z-score anyways.
[] Independent signals in high LD region.
[] Small size yet low purity.
[] Show susie iteration gradual convergence.
- Look in N = 2 simulations of small CS size yet large number of iterations.
[] Show why susie can pick up one variable twice.
[] Show why susie captures two same signals in two iterations.

[X] A vignette showing finemapping analysis with other methods and compare PIP results with susie
[] Comparison of susie iterations with step-wise conditional regression.
[] Comparison of susie with SER. This would be a special case of above.

Issues in reporting independent signals: duplicated? near by clusters? Should assess correlation between CS before reporting and how?
How does multiple-testing come into play here?
How does the prior setting actually reflects prior on number of causal as FINEMAP does it?

Revision	Author	Date	Message
e14a9ec	Gao Wang	2018-06-22	Reorgnize size vs purity plot layout
ce631b9	Gao Wang	2018-06-22	Add discussion & exploritory items