A Simple Distribution-Free Algorithm for Generating Simulated High-Dimensional Correlated Data with an Autoregressive...

Area of Science:

• Statistics
• Computational Biology
• Data Science

Background:

• Generating high-dimensional sequences with autoregressive structures is challenging.
• Existing methods often require large correlation matrices, limiting scalability.
• Quantile correlation is a key measure for dependence in such sequences.

Purpose of the Study:

• To present a distribution-free method for generating high-dimensional autoregressive sequences.
• To utilize quantile correlation as the primary dependence measure.
• To develop an algorithm that avoids defining a single large correlation matrix.

Main Methods:

• The proposed method breaks down the high-dimensional sequence into smaller, manageable segments.
• It uses quantile correlation to define dependence between contiguous variables.
• Simulations were performed using proteomics data to validate the approach.

Main Results:

• The algorithm successfully generates high-dimensional sequences with autoregressive properties.
• Splitting the sequence generation into smaller parts resulted in negligible loss of quantile correlation.
• The method proved effective when applied to proteomics datasets.

Conclusions:

• The developed distribution-free method offers an efficient way to generate complex, high-dimensional data.
• Breaking sequences into parts is a viable strategy to manage computational complexity without significant loss of dependence information.
• This approach has potential applications in fields like bioinformatics and statistical modeling.