findSpectrum: Determine number of informative latent dimensions

Description Usage Arguments Details Value Author(s) See Also Examples

Description

Identifies the dimensionality of the latent space

Usage

1
findSpectrum(x, frac = 100)

Arguments

x

A numeric vector with eigenvalues

frac

Fraction or number (if frac > 1) of eigengaps used to perform linear fit. (default: 100)

Details

Similar to a scree plot, this method generates a simple line segement plot showing the lagged differences between ordered eigenvalues (eigengaps). A linear fit is calucated on a fraction of top ranked values to identify informative eigenvectors.

Value

A numeric vector with indices of relevant dimensions

Author(s)

Daniel C. Ellwanger

See Also

pca embedSamples

Examples

1
2
3
4
5
6
7
8
9
# Example data
data(exSCE)

# Embedding
res <- embedSamples(exSCE)

# Find spectrum
d <- findSpectrum(res$eigenvalues, frac=30)
d

CellTrails documentation built on Nov. 8, 2020, 5:53 p.m.