We have observed lately that when performing PLSDA, there are some cases in which t2>t1. What could this mean?
It is worth mentioning that the stats on some of the datasets are not always significant, but they are in the mentioned cases in which t2>t1.
Thank you in advance for your help!
The PLS-DA algorithm computes the components by decreased covariance between X and the response y. On the score plots, however, are displayed the proportion of the X variance captured by the components as t1, t2, etc. So it might happen the the variance is increased although the overall covariance is decreased. In any case, it is of critical importance to consider only models with a significant pQ2 value.