testClusters {BiSeq} | R Documentation |
CpG clusters are tested with a cluster-wise FDR level.
testClusters(locCor, FDR.cluster)
locCor |
Output of |
FDR.cluster |
A |
CpG clusters containing at least one differentially methylated location are detected.
A list is returned:
FDR.cluster |
Chosen WFDR (weighted FDR) for clusters. |
CpGs.clust.reject |
A |
CpGs.clust.not.reject |
A |
clusters.reject |
A |
clusters.not.reject |
A |
sigma.clusters.reject |
The standard deviations for z-scores within each rejected cluster. |
variogram |
The variogram matrix. |
m |
Number of clusters tested. |
k |
Number of clusters rejected. |
u.1 |
Cutoff point of the largest P value rejected. |
Katja Hebestreit
Yoav Benjamini and Ruth Heller (2007): False Discovery Rates for Spatial Signals. American Statistical Association, 102 (480): 1272-81.
## Variogram under Null hypothesis (for resampled data): data(vario) plot(vario$variogram$v) vario.sm <- smoothVariogram(vario, sill=0.9) # auxiliary object to get the pValsList for the test # results of interest: data(betaResults) vario.aux <- makeVariogram(betaResults, make.variogram=FALSE) # Replace the pValsList slot: vario.sm$pValsList <- vario.aux$pValsList ## vario.sm contains the smoothed variogram under the Null hypothesis as ## well as the p Values that the group has an effect on DNA methylation. locCor <- estLocCor(vario.sm) clusters.rej <- testClusters(locCor, FDR.cluster = 0.1)