summarizeGroupedDirection: Summarize overall direction of grouped tests

In LTLA/metapod: Meta-Analyses on P-Values of Differential Analyses

Summarize overall direction of grouped tests

Description

Summarize the overall direction of grouped tests in a meta-analysis, based on the influential tests defined by one of the grouped* functions.

Usage

summarizeGroupedDirection(effects, grouping, influential, threshold = 0)

Arguments

effects	A numeric vector containing the effect size for each test.
grouping	A vector of factor of length equal to effects, specifying the assigned group for each tests. Alternatively, an rle object where each run corresponds to a group and specifies the entries of effects belonging to that group. This assumes that effects is ordered such that all entries in the same group are adjacent to each other.
influential	A logical vector of length equal to effects, indicating whether each test is influential in its assigned group.
threshold	Numeric scalar defining the threshold at which an effect is “"up"” or “"down"”.

Details

We focus on the direction of effect for the influential tests that actually contribute to a group's final p-value. For example, if we did our meta-analysis using parallelSimes,we are not particularly concerned about the direction of tests with large p-values. Thus, we just ignore them when summarizing the group's direction in this function. Otherwise, we would unnecessarily obtain a mixed direction of effect if a test with a large p-value had a weakly opposing effect.

Of course, the interpretation of “influential” really depends on the choice of meta-analysis strategy. It is also possible that this function reports a single direction when the group really is mixed, e.g., if the tests with the lowest p-values are changing in one direction but tests with weaker but still interesting effects are changing in the other direction. The extent to which this is of interest is left to the discretion of the user.

Value

A character vector of length equal to the number of groups. Each entry can be:

• "up", if all influential tests have effects above threshold.

• "down", if all influential tests have effects below threshold.

• "none", if all influential tests have effects equal to threshold.

• "mixed", if there are influential tests with effects above and below threshold.

Author(s)

Aaron Lun

See Also

groupedSimes and related grouped* functions, to obtain influential.

summarizeParallelDirection, for the equivalent function based on parallel tests.

countGroupedDirection, to count the number of effects in each direction.

Examples

p <- rbeta(100, 0.5, 1)
eff <- rnorm(100)
g <- sample(20, 100, replace=TRUE)

out <- groupedSimes(p, g)
(dir <- summarizeGroupedDirection(eff, g, out$influential))

LTLA/metapod documentation built on Jan. 19, 2024, 11:49 p.m.