PLSDA variates are different after updating from version 6.13 to 6.3.1

[Former user]

Hello,

I have recently updated MixOmics to version 6.3.1 from version 6.1.3. I have noticed that in the newest MixOmics version, my first PLSDA component is mirrored in PlotIndiv plots compared to what I had before. I have attached my dataset, an R script and Plotindiv plots from both MixOmics versions. I have also included my sessionInfo as a txt file.

What is the reason for the change, and which MixOmics version produces the correct result?

Thank you in advance,

Andrew

[Kim-Anh Le Cao]

Dear Andrew,

Thank you for contacting us. What you report (from what I see from your plots) is simply the sign of the variate 1 that have been inverted (from positive to negative, or vice versa), which is common for this type of method. It does not affect the interpretation of the sample plots in your case. One of the reason might be that we had to fix the algorithm for NA values and that may have swapped the variates values for some particular data.

Also note that in the NEWS file (availble on CRAN in the materials: https://cran.r-project.org/web/packages/mixOmics/index.html) you will see the list of changes between the version.

Thanks for using mixOmics, there are more examples on our website www.mixOmics.org if you need.

Kim-Anh

[Kim-Anh Le Cao]

Swap of signs between version, as explained.