Skip to content

Latest commit

 

History

History
108 lines (77 loc) · 4.15 KB

README.md

File metadata and controls

108 lines (77 loc) · 4.15 KB

RoLDSIS – Regression on Low-Dimension Spanned Input Space

Introduction

The present repository contains the data and the code for analysing the data related to a publication involving the RoLDSIS method. RoLDSIS stands for “Regression on Low-Dimension Spanned Input Space”. The data for an electroencephalography (EEG) experiment involving phonemic identification is included in the repository. The scripts in the R language, for processing and analysing the data, as well as producing the figures that appear in the following article:

Santana, A., Barbosa A., Yehia H., and Laboissière R. (2021) A dimension reduction technique applied to regression on high dimension, low sample size neurophysiological data sets. BMC Neurosci 22(1). DOI: 10.1186/s12868-020-00605-0.

Installation

The code can be obtained with the following command:

git clone https://github.com/RoLDSIS/code RoLDSIS

N.B.: Developers with write access right to the repository should do, instead:

git clone git@github.com:RoLDSIS/code.git RoLDSIS

Running the analysis and producing the figures

The system packages R, make, and pdftk are needed. The necessary R packages are installed as required by the scripts.

In GNU/Linux systems, the analysis of the data and the generation of the figures are done with the following commands :

cd RoLDSIS/script
make

Description of the contents of this repository

Scripts containing supporting functions and general parameters definition

RoLDSIS functions

Cross-validation for RolDSIS, LASSO, Ridge Regression and SPLS

Generation of figures

Data

Description of the generated figures:

  • psy-VOT-Snn.pdf: Results of phonemic identification experiment for subject nn
  • erp-VOT-Ativo-Snn.pdf: Average ERPs for the five stimuli for subject nn
  • cv-direction-Snn.pdf: RoldSIS result represented in scalogram for subject nn
  • cv-projections-Snn.pdf: RoLDSIS projected responses for subject nn
  • cv-errors.pdf: Cross-validation errors for RoLDSIS, LASSO, Ridge Regression, and SPLS
  • cv-scalograms.pdf: Histograms for the regression methods of regressed coefficients
  • trials-observation.pdf: RMS prediction error for all subjects using different number of averaged points

Documentation

Licensing conditions

The files in this repository are made available under the conditions of the GNU Public License, version 3 or later. No warranties. The RolDSIS has been submitted for publication. If you use the software or the data of this repository, please give credit.

Authors