pip: Compute posterior inclusion probabilities (PIPs)

In jackstraw: Statistical Inference for Unsupervised Learning

Compute posterior inclusion probabilities (PIPs)

Description

From a set of p-values, computes posterior probabilities that a feature should be truly included. For example, membership inclusion in a given cluster can be improved by filtering low quality members. In using PCA and related methods, it helps select variables that are truly associated with given latent variables.

Usage

pip(pvalue, group = NULL, pi0 = NULL, verbose = TRUE, ...)

Arguments

pvalue	a vector of p-values.
group	a vector of group indicators (optional). If provided, PIP analysis is stratified. Assumes groups are in 1:k where k is the number of unique groups.
pi0	a vector of pi0 values (optional). Its length has to be either 1 or equal the number of groups.
verbose	If TRUE, reports information.
...	optional arguments for lfdr to control a local FDR estimation.

Details

This function requires the Bioconductor qvalue package to be installed.

Value

pip returns a vector of posterior inclusion probabilities

Author(s)

Neo Christopher Chung [e-mail] John R. Yamamoto-Wilson

References

Chung (2020) Statistical significance of cluster membership for unsupervised evaluation of cell identities. Bioinformatics, 36(10): 3107–3114 \Sexpr[results=rd]{tools:::Rd_expr_doi("10.1093/bioinformatics/btaa087")}

Chung (2014) "Jackstraw Weighted Shrinkage for Principal Component Analysis and Covariance Matrix" in Statistical Inference of Variables Driving Systematic Variation in High-Dimensional Biological Data. PhD thesis, Princeton University. https://www.proquest.com/openview/e90b562d689cf3a021c35a93c6f346db/1?pq-origsite=gscholar&cbl=18750

jackstraw documentation built on Sept. 17, 2024, 1:07 a.m.