pars_start: pars_start

In jamesliley/subtest: Test whether two putative subgroups of a disease phenotype show evidence of differential genetic basis.

Description

Obtain a set of starting parameters for E-M algorithm, given Z_a and Z_d scores and LDAK weights. Because of the potential for local maxima in the likeilhood landscape, it is important to start the algorithm at several points. Since the E-M algorithm is computationally intensive, it is useful to start as close as possible to the actual maxima, and at as few points as possible. This function seeks a small number of 'promising' start points with high likelihood, sufficiently far from each other.

Usage

pars_start(Z, weights = rep(1, dim(Z)[1]), H = 1, n1 = 100, n2 = 10,
  n3 = 5, C = 1, nx = 3, seed = NULL)

Arguments

Z	an n x 2 array; Z[i,1], Z[i,2] are the Z_d and Z_a scores respectively for the ith SNP
weights	SNP weights to adjust for LD; output from LDAK procedure
H	hypothesis, 0 or 1
n1	begin with n1 random parameter sets from pars_rand, chosen according to prior distributions of parameters (see documentation for pars_rand)
n2	trim initial list to this many well-separated sets of parameters; each of these parameter sets undergoes nx steps of the EM algorithm
n3	finally take n3 well-separated sets of parameters for entry into the final E-M algorithm
C	scaling factor for adjustment
nx	use this many iterations of the E-M algorithm on each of the n2 sets of paramaters
seed	random seed for generating results. Use to regenerate.

Author(s)

James Liley

jamesliley/subtest documentation built on May 18, 2019, 11:21 a.m.