Variance of the VIP statistics from PLS

Hi everyone,

I have performed PLS regression using the MixOmics package, and I am able to obtain VIP statistics for the predictor variables (X variables, independent variables).

However, I am wondering if it is possible to obtain a measure of the variance of these statistics using MixOmics? E.g., a standard error or a confidence interval for the VIP statistics?

Please let me know if any of you have ad experience with this.

Thank you!

hi @bort

Thank you for your question. The VIP is a measure for each variable, not a statistic. You would need to use cross-validation and extract the VIP for each fold if you would like to do something more sophisticated (we have not).

Below is a screenshot of the handbook draft (disregard my blue notes!) about how it is calculated:


Hi Kim-Anh,

Thank you very much for taking the time to respond to my inquiry, I sincerely appreciate it!

This makes sense, as I have only ever seen 95% CIs for VIPs obtained via jack-knife resampling. Perhaps that can be something to consider adding in future iterations? Just a thought!

And thank you for sharing this information on how the VIP is calculated. It is also very helpful to me.

All the best!