README:

The program PrivPick.py implements various methods for picking high scoring SNPs in a differentially private manner.

The methods are the laplacian based method and exponential mechanism based method for picking high scoring SNPs introduced by Uhler et al, where the score is based off the allelic test statistic. We use both the neighbor distance based method (our main contribution is to give an algorithm that allows the calculation of neighbor distance faster) and the naive score based methods.

Each of our methods takes 4 inputs (in the following order):
-mret, the number of SNPs to return,
-epsilon, the privacy parameter,
-r, an array of three tuples, were r[i]=[r0,r1,r2], and r0,r1,r2 are the number of cases with 0, 1 or 2 version of the minor allele,
-s, an array of three tuples, were s[i]=[s0,s1,s2], and s0,s1,s2 are the number of controls with 0, 1 or 2 version of the minor allele,
They are ordered 

Each of the methods returns a list of mret indices, the indices of the mret highest scoring SNPs. These methods are:
-scoreLap which uses the Laplacian method on the allelic test statistic
-scoreExp, uses the exponential mechanism on the allelic test statistic
-neighExp, uses the exponential mechanism on the neighbor distance
-neighLap, uses the laplacian method on the neighbor distance

The rest of the methods are helper methods to these four methods. Questions or issues can be sent to seanken@mit.edu.