Umpire (version 2.0.11)

ClinicalNoiseModel: A Noise Model for Clinical Data

Description

A ClinicalNoiseModel represents the additional human and measurement noise that is layered on top of any biological variabilty when measuring clinical variables.

Usage

ClinicalNoiseModel(nFeatures, shape = 1.02, scale = 0.05/shape)

Value

An object of class NoiseModel.

Arguments

nFeatures

An integer; the number of additive scale parameters to sample from the gamma distribution.

shape

The shape gamma hyperparameter describing the standard deviation of additive noise.

scale

The scale gamma hyperparameter describing the standard deviation of additive noise.

Author

Kevin R. Coombes , Caitlin E. Coombes

Details

We model both additive and multiplicative noise, so that the observed expression of clinical variable c in sample i is given by: \(Y_ci = S_ci + E_ci\), where Y_ci = observed expression, S_ci = true biological signal. In the ClinicalNoiseModel (as opposed to the NoiseModel), we model the additive noise as \(E_ci ~ N(0,tau)\), without multiplicative noise or an additive bias/offset in the clinical model. The standard deviation hyperparameters of the additive noise tau is modeled by the gamma distribution \(tau ~ Gamma(shape, scale)\)

See Also

NoiseModel-class,

Examples

showClass("NoiseModel")

## generate a ClinicalEngine with 20 features and 4 clusters
ce <- ClinicalEngine(20, 4, TRUE)
## generate 300 simulated patients
set.seed(194718)
dset <- rand(ce, 300)

cnm <- ClinicalNoiseModel(nrow(ce@localenv$eng), shape=2, scale=0.1)
cnm

noisy <- blur(cnm, dset$data)
hist(noisy)