selectTop: Selects the top genomic events, based on posterior...

In baySeq: Empirical Bayesian analysis of patterns of differential expression in count data

Description

This function subsets a countData object by selecting those events that best (or least) represent a model, based on the posterior likelihoods estimated for that model and some threshold. Selection can be done for a specific model (and ordering of the data under that model) or for all models (and all orderings).

Usage

selectTop(cD, group, ordering, orderings = TRUE, decreasing = TRUE,
number = 10, likelihood, FDR, FWER, posteriors)

Arguments

cD	A countData object, with a populated ‘@posteriors’ slot (unless ‘posteriors’ is specified; see below).
group	Optionally, the model of interest, as defined in the ‘@groups’ slot of the countData object. If unspecified, subsets for all models will be returned as a list.
ordering	If ‘group’ is specified, a particular ordering of the data based on that group can also be specified.
orderings	If no group is specified, should the selection of models also be split by the orderings of the data under the models? Defaults to TRUE.
decreasing	If FALSE, considers the data with the lowest posterior likelihoods, rather than the greatest (i.e., selects those data least likely to conform to a particular model.
number	If given, selects the top 'number' of genomic events for each model (and optionally, ordering). Ignored if another selection criteria is chosen, unless this criteria would return no values.
likelihood	If given, selects all genomic events for a given model (and optionally, ordering) with posterior likelihood exceeding this value.
FDR	If given, selects all genomic events for a given model (and optionally, ordering) with false discovery rate below this value. Ignored if likelihood is specified.
FWER	If given, selects all genomic events for a given model (and optionally, ordering) with family-wise error rate below this value. Ignored if likelihood or FDR is specified.
posteriors	If given, a vector of log-posterior likelihoods to use instead of those in the ‘@posteriors’ slot of the ‘cD’ object.

Value

Either a single countData object (if ‘group’ is specified), or a named list of countData objects.

Author(s)

Thomas J. Hardcastle

See Also

topCounts

Examples

# We load in a `countData' object containing the estimated posterior
# likelihoods of expression (see `getLikelihoods').

data(CDPost)

# select from all models and orderings with FDR equal to or lower than 0.01.

selectTop(CDPost, FDR = 0.01)

baySeq documentation built on Nov. 8, 2020, 5:43 p.m.