clustermodelr: Model clustered, correlated data

In brentp/clustermodelr: Models on clusters of correlated data

Description

clustermodelr provides a consistent, simple interface to model correlated data using a number of different methds:

GEE:

Generalized Estimating Equations with all correlation structures available from geepack. geer

mixed-effect model:

Mixed effect model in lme4 syntax mixed_modelr

combiner:

Calculates the p-value for each entry in the cluster then combines the p-values adjusting for correlation with either stouffer_liptak.combine or zscore.combine

bumping:

something like bump-hunting but takes a putative "bump" and repeatedly compares coefficients of estimated covariates to the observed to assign significance. bumpingr

SKAT:

SKAT already accepts a matrix to test a null model. This just provides an interface that matches the rest of the functions in this package skatr

Details

Each of these functions will accept a formula like:

methylation ~ disease + age

(with a random intercept for mixed_modelr) where methylation need not be methylation values, but is assumed to be a matrix of correlated values.

For each of these functions, the return value will be a vector of:

c(covariate, p, coef.estimate)

where the covariate is taken as the first element on the RHS of the formula so disease in the formula above.

brentp/clustermodelr documentation built on May 13, 2019, 5:11 a.m.