plrs.test: Likelihood ratio test for a 'plrs' model

In plrs: Piecewise Linear Regression Splines (PLRS) for the association between DNA copy number and gene expression

Description

Test whether copy number has an effect on mRNA expression.

Usage

plrs.test(object, alpha=0.05)

Arguments

object	An object of class plrs-class
alpha	Significance level

Details

Two cases present themselves:

1. The model is unconstrained. Thus, the model under the null hypothesis is the intercept and an F-test is performed.

2. The model is constrained and the following hypothesis are tested:
H0: All constraints are actives (=)
H1: At least one constraint is strict (>)
Under H0, we always have the intercept model. Indeed, if constr.slopes = 1 (or 2) and constr.intercepts = T, then the only parameter free of inequality constraint is the overall intercept. If constr.intercepts = F, the local intercepts are additionnaly constrained to be 0 in order to obtain the intercept model under the null. The likelihood ratio statistic (unknown variance) is asymptotically distributed as a weighted mixture of Beta distribution (cf Gromping (2010)). Calculation of p-values is based on functions ic.weights and pbetabar of package ic.infer. The package mvtnorm is also involved.

In both cases the input model is taken as the model under the alternative.

Value

A list object with the following components:

stat	Test statistic
pvalue	Calculated pvalue
wt.bar	Weights (if the model is constrained)
df.bar	Degrees of freedom.
unconstr	Unconstrained model of class plrs-class
qbetabar	(1-alpha) quantile of the beta mixture distribution
alpha	Significance level

Author(s)

Gwenael G.R. Leday [e-mail]

References

Gromping, U. (2010). Inference with linear equality and inequality constraints using R: The package ic.infer. J Stat Softw, 33(i10).

Examples

# Simulate data
sim <- plrs.sim(n=80, states=2, sigma=0.5)

# Fit a model 
model <- plrs(expr=sim$expr, cghseg=sim$seg, cghcall=sim$cal)

# Testing
model <- plrs.test(model)
model

plrs documentation built on April 28, 2020, 6:09 p.m.