In the SRBCT Permuted Labels Case Study, cross-validated error rate is listed as the best way to assess if data are appropriate for sPLS-DA. Output is displayed for “Normal model performance metrics” vs. “Permuted model performance metrics”.
Can you please share the code/command used to generate those output tables for performance metrics?
If it’s possible to share all of the code for this case study instead of only outputs–similar to the other case studies–that would be very helpful. Especially the initial permutation of class labels step to ensure mixOmics users are following along correctly. Thanks for your consideration!
Hi @StatsQs,
Thank you for the feedback, I agree that having the code for the case studies makes them much more useful. I’ve updated the webpage so it now shows the code to permute the class labels and create the models. At the top of the page I’ve also now included a link to the full R script with all of the code similar what we have for the other case studies.
Note that we have a bug in the current Bioconductor release of mixOmics which means the ellipses from plotIndiv() are the wrong colours, if you would like to reproduce the plots as they are on the webpage you can install the latest mixOmics version from Github using BiocManager::install('mixOmicsTeam/mixOmics'), for the rest of the code the stable Bioconductor mixOmics version should be fine!
Cheers,
Eva