dispersion: Calculate dispersion metrics

In UCLA-BD2K/CV.Signature.TCP: Cardiovascular Signature Temporal Clustering Platform

Description

For each variable (row), calculate a dispersion (var/mean) and a z-score of normalized dispersion. A z-score of normalized dispersion is motivated due to correlation between means and variances. Variables are binned (by default,n_bins=10) by their means. Within a bin, a z-score is calculated. It's recommended that you look at both results zdisp and disp to decide how to filter the data.

Usage

dispersion(
  dat,
  n_bins = 10,
  center.dat = TRUE,
  scale.dat = FALSE,
  verbose = FALSE,
  seed = NULL,
  ...
)

Arguments

dat	a time-series data matrix with m biomarkers as rows, over n time points (columns).
center.dat	a logical specifying to center the input and denoised data. By default, TRUE.
scale.dat	a logical specifying to scale the input and denoised data. By default, FALSE.
verbose	a logical specifying to print the computational progress. By default, FALSE.
seed	a seed for the random number generator.
...	optional arguments.
timepoints	a vector of time points for columns of dat.
dof	the degree of freedom in spline.smooth. By default, cv performs cross-validation per-variable.

Value

dispersion returns a data.frame of statistics for m variables (rows).

Author(s)

Neo Christopher Chung [e-mail]

Examples

## Not run: 
data(cys_optm)
meta <- cys_optm[,1:4]
optm <- log(cys_optm[meta$Select,5:10])
optm <- t(scale(t(optm), scale=TRUE, center=TRUE))
days <- as.numeric(colnames(optm))

disp_optm <- dispersion(optm, timepoints = days, dof="cv")
disp_optm <- cbind(meta, disp_optm)
# make a histogram of dispersion statistics
hist(disp_optm$disp, 100)
# make a histogram of z-score of normalized dispersion
hist(disp_optm$zdisp, 100)
# library(readr)
# write_excel_csv(disp_optm, file="~/coptm_dispersion.csv")

## End(Not run)

UCLA-BD2K/CV.Signature.TCP documentation built on May 15, 2020, 11:27 p.m.