calculateJB

We measure non-Gaussianity using Jarque-Bera (JB) statistic, which is a weighted combination of squared skewness and kurtosis, <a href="https://www.jstor.org/stable/1403192?origin=crossref">JB paper</a>.
The data has to be standardized and mean 0 and sd to 1.

Implementation of SING algorithm to extract joint and individual non-Gaussian components from two datasets. SING uses an objective function that maximizes the skewness and kurtosis of latent components with a penalty to enhance the similarity between subject scores. Unlike other existing methods, SING does not use PCA for dimension reduction, but rather uses non-Gaussianity, which can improve feature extraction. Benjamin B.Risk, Irina Gaynanova (2021) <doi:10.1214/21-AOAS1466>.

Liangkang Wang

singR

Simultaneous Non-Gaussian Component Analysis

Irina Gaynanova

Benjamin Risk

calculateJB function

<dl><dt>S</dt>
<dd>the variable loadings r x px.</dd>
<dt>U</dt>
<dd>U matrix for matched columns rj x n</dd>
<dt>X</dt>
<dd>whitened data matrix n x px, data = whitenerXA %*% dXcentered</dd>
<dt>alpha</dt>
<dd>JB weighting of skewness and kurtosis. default = 0.8</dd></dl>

Arguments

We measure non-Gaussianity using Jarque-Bera (JB) statistic, which is a weighted combination of squared skewness and kurtosis, <a href='https://www.jstor.org/stable/1403192?origin=crossref'>JB paper</a>.
The data has to be standardized and mean 0 and sd to 1.

Calculates the sum of the JB scores across all components, useful for determining rho. — calculateJB

<dl>

<dt>S</dt>
<dd>the variable loadings r x px.</dd>


<dt>U</dt>
<dd>U matrix for matched columns rj x n</dd>


<dt>X</dt>
<dd>whitened data matrix n x px, data = whitenerXA %*% dXcentered</dd>


<dt>alpha</dt>
<dd>JB weighting of skewness and kurtosis. default = 0.8</dd>

</dl>

calculateJB: Calculates the sum of the JB scores across all components, useful for determining rho.

Description

Usage

Value

Arguments