BiSeq (version 1.12.0)

makeVariogram: Variogram estimator.

Description

A function which estimates the variogram of the z-scores in the given data frame.

Usage

makeVariogram(test.out, make.variogram, sample.clusters, max.dist)

Arguments

test.out

A data.frame. Usually the output of betaRegression. Must contain columns chr, pos, p.val and cluster.id.

make.variogram

A logical. Default is TRUE.

sample.clusters

Can speed up variogram estimation significantly. Default is NULL, and all data is used to estimate the variogram. If set to numeric, the variogram will be estimated on the basis of the data of randomly selected sample.clusters only. Especially useful if there are many clusters.

max.dist

Can speed up variogram estimation significantly. The variogram is estimated for distances until this threshold. Default is 500 base pairs, since the variogram usually does not change for distances larger than 100 base pairs, because methylation of CpG sites further away are not correlated anymore. Especially useful if there are large clusters.

Value

• A list:
• variogram
  A list of two: A matrix, called v with columns h and v, and a numeric, called h.est. v comprises the data that was used to estimate the variogram. h.est comprises the distances seen in the data. If sample.clusters=NULL, h.est is identical to v$h.
• pValsList
  A list of data frames. Each data frame corresponds to a CpG cluster and contains same information as test.out plus the columns z.score and pos.new (position corresponding to the respective CpG cluster).

Details

For each CpG site the z-score is determined by qnorm(1 - P value). The variogram of the z-scores of locations $k$ and $l$ within one cluster is estimated robustly by $$2 \hat{\gamma}(h) = [median{(Z_{k}-Z_{l})^2: (s_{k}, s_{l}) \in N(h)}] / .455$$.

References

Yoav Benjamini and Ruth Heller (2007): False Discovery Rates for Spatial Signals. American Statistical Association, 102 (480): 1272-81.

See Also

betaRegression

Examples

data(betaResults)

vario <- makeVariogram(betaResults)

plot(vario$variogram$v)