elbow_method: Runs elbow method

In VingronLab/APL: Association Plots

Runs elbow method

Description

Helper function for pick_dims() to run the elbow method.

Usage

elbow_method(obj, mat, reps, python = FALSE, return_plot = FALSE)

Arguments

obj	A "cacomp" object as outputted from 'cacomp()'
mat	A numeric matrix. For sequencing a count matrix, gene expression values with genes in rows and samples/cells in columns. Should contain row and column names.
reps	Integer. Number of permutations to perform when choosing "elbow_rule".
python	A logical value indicating whether to use singular value decomposition from the python package torch. This implementation dramatically speeds up computation compared to 'svd()' in R.
return_plot	TRUE/FALSE. Whether a plot should be returned when choosing "elbow_rule".

Value

'elbow_method' (for 'return_plot=TRUE') returns a list with two elements: "dims" contains the number of dimensions and "plot" a ggplot. if 'return_plot=TRUE' it just returns the number of picked dimensions.

References

Ciampi, Antonio, González Marcos, Ana and Castejón Limas, Manuel.
Correspondence analysis and 2-way clustering. (2005), SORT 29(1).

Examples

# Get example data from Seurat
library(Seurat)
set.seed(2358)
cnts <- as.matrix(Seurat::GetAssayData(pbmc_small,
                                       assay = "RNA",
                                       slot = "data"))
# Run correspondence analysis.
ca <- cacomp(obj = cnts)

# pick dimensions with the elbow rule. Returns list.
pd <- pick_dims(obj = ca,
                mat = cnts,
                method = "elbow_rule",
                return_plot = TRUE,
                reps = 10)
pd$plot
ca_sub <- subset_dims(ca, dims = pd$dims)

VingronLab/APL documentation built on Feb. 14, 2024, 11:40 a.m.