Fits a mixture of Gaussian to a set of one dimensional points.

Share:

Description

This is the workhorse function, essentially an R wrapper around a lot of C code. It fits GLM models to the data.

Usage

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
CNV.fitModel(ncomp, 
	     nind, 
	     hyp = "H0", 
	     data, 
	     logit.offset,
             design.matrix.mean, 
	     design.matrix.variance,
             design.matrix.disease, 
	     pi.model = 0, 
	     mix.model = 10,
             control = list(tol = 1e-05, max.iter = 3000, min.freq= 4))

Arguments

ncomp

integer, number of components to fit to the data

nind

integer, total number of data points

hyp

Hypothesis, can be either H0 or H1

data

The data frame containing the data, in an expanded form (one point per individual and copy number)

logit.offset

An option most users will not use. It sets an offset when fitting the logit model for the disease status. This is used to obtain a profile likelihood when the disease parameter beta varies.

design.matrix.mean

The design matrix that relate mean cluster locations with batch.copy numbers.

design.matrix.variance

The design matrix for the cluster variances.

design.matrix.disease

The design matrix for the disease model.

pi.model

0,1,2 fit disease, hetero and quantitative models respectively.

mix.model

Specifies model for the components.

control

A list of parameters that control the behavior of the fitting.

Details

The user is very unlikely to actually use that function which is meant as an internal routine, a wrapper around the C code of the package. This function is called by the more user friendly function CNVtest.binary.

Value

data

The input expanded data frame, but with the posterior probabilities estimated.

status

A marker of convergence

Author(s)

Vincent Plagnol vincent.plagnol@cimr.cam.ac.uk and Chris Barnes christopher.barnes@imperial.ac.uk

See Also

CNVtest.binary