haplofreq: Fit Hardy-Weinberg equilibrium model for haplotype...

In HaploSurvival: Modelling of haplo-effects for survival data, including competing risks. Effect of haplomatch for BMT studies.

Description

Fit MLE of Hardy-Weinberg equilibrium model for haplotype frequencies based on genotype data

Input is genotype data.

The haplotype model assumes Hardy-Weinberg equilibrium

P(H=(h_k,h_j)) = π_k pi_j

with K different haplotypes, where we parametrize the frequencies as

π_k= \frac{ \exp(θ_k) }{\exp(θ_1)+...+\exp(θ_{K-1})+1}

We allow a regression design on the haplotype parameters to reduce the dimensionality

θ = G α

where G is matrix of dimension (K-1) x (d-1). The α and θ version both have the same baseline that is a single haplotype.

Usage

haplo.freqs(geno.type, geno.setup=NULL, Nit = 10,detail = 0,
haplo.freq = NULL, step = 1,lev.marq=1,min.lev.marq=0, 
haplo.design=NULL,haplo.baseline=NULL,alpha=NULL)

Arguments

geno.type	a data.frame with the variables.
geno.setup	setup form genotype data.
Nit	number of iterations for Newton-Raphson algorithm.
detail	if 0 no details is printed during iterations, if 1 details are given.
haplo.freq	haplo type frequencies for starting values or for fixed values.
step	step size for iteration steps.
lev.marq	starting value for Levenberg-Marquart algorithm.
min.lev.marq	stopping value for Levenberg-Marquart algorithm after initial steps, min.lev.marq=0 is stopping value for Levenberg-Marquart algorithm after initial steps, min.lev.marq=0 is Newton-Raphson steps, that are needed to get correct standard errors.
haplo.design	design for haplo parameters, default is diagonal of size K-1
haplo.baseline	gives the name of the geno.setup that should be used as baseline, default is last haplotype.
alpha	starting value of parameters for haplo.design

Value

returns an object of type "haplo.freq". With the following arguments:

haplo.alpha	MLE of alpha parameters for haplo.design.
haplo.pars	MLE of haplotype parameters.
haplo.freq	MLE of haplotype frequencies.
var.haplo.alpha	variance matrix for alpha
D2linv	second derivative of log-likelihood.
score	score for likelihood.
haplo.design	design that relates haplotype parameters to alpha parameters.
alpha.iid	iid decomposition of alpha parameters, in this case score contributions for each subject for score of for log-likelihood.
freq.names	names of haplotypes

Author(s)

Thomas Scheike and Jeremy Silver

References

Scheike, Martinussen and Silver, Haplotype effects for survival data, Submitted.

Examples

n<-200
g<-matrix(rbinom(200*4,1,0.5),n,4); # simple genotypes

gs<-geno.setup(g); 

hgs<-haplo.freqs(g,geno.setup=gs)
summary(hgs); 
hgs$haplo.freq         # haplo-type frequencies
hgs$haplo.pars         #  related parameters 
hgs$haplo.alpha        #  related parameters  haplo.pars = X alpha
hgs$var.haplo.alpha    #  variance of alpha parameters 

# structured haplo-frequency model, baseline is set to "0,0"

gs<-geno.setup(g,haplo.baseline="0,0"); 

X<-matrix(1,3,1); 
hgs<-haplo.freqs(g,geno.setup=gs,haplo.design=X)
summary(hgs); 
hgs$haplo.freq         # haplo-type frequencies
hgs$haplo.pars         # related parameters 
hgs$haplo.alpha        # related parameters  haplo.pars = X alpha
hgs$var.haplo.alpha    # variance of alpha parameters 

HaploSurvival documentation built on May 2, 2019, 5:49 p.m.