mincLogLikRatioParametricBootstrap: computes parametric bootstrap on mincLogLikRatio output

Description Usage Arguments Value

View source: R/minc_lmer.R

Description

The Log Likelihood Ratio tests closely approximates a Chi-squared distribution when the number of groups (i.e. individual subjects in a longitudinal study) is large (>50), but can be anticonservative when small. A parametric bootstrap test, in which data is randomly simulated from the null model and then fit with both models, can give the correct p-value. Here we compute the parametric boostrap on a small number of randomly chosen voxels to get a sense of biased the estimated p-values from the log likelihood ratio test really were.

Usage

1
2
mincLogLikRatioParametricBootstrap(logLikOutput, selection = "random",
  nsims = 500, nvoxels = 50)

Arguments

logLikOutput

the output from mincLogLikRatio

selection

the algorithm for randomly chosing voxels. Only "random" works for now.

nsims

the number of simulations to run per voxel

nvoxels

the number of voxels to run the parametric bootstrap on

Value

a matrix containing the chi-square p-values and the bootstrapped p-values


Mouse-Imaging-Centre/RMINC documentation built on Oct. 5, 2018, 9:23 a.m.