### Simple Method

(*See *`just_ave`

directory on hpc)

Now that we know that a GAM is not being fitted, our method is very simple.

We estimate the true effect at the causal variant and then simulate many credible sets from the same distribution. We then find the proportion of these which contain the true causal variant. We do this for each variant considered causal and weight each proportion by that variantâ€™s posterior probability.

I am investigating how fast I can get this method. The slow step is forming credible sets. Chris has wrote a quicker function in Cpp to do this, which I am now using.

The results below suggest that the correction doesnâ€™t change much between using 1000 (~55 secs) or 10,000 (~8 mins) simulated credible sets. Even the error bars donâ€™t change much (???)