FAQ  Frequently Asked Questions
Issue:

Why is the confidence limit on scores different in PLS_Toolbox after version 3.5.2?
Possible Solutions:

Prior to Version 3.5.2, PLS_Toolbox used a limit calculation for scores which was based on a ttest calculation for the individual PC (or PCs) you were viewing. Although this is based on sound statistics, the problem was that the approach tended to underestimate the limit and was not consistent with the limit calculation for Hotelling's T^{2} (the limit of which is actually based on an ftest).
As of version 3.5.2, the individual score limits are now based on the Hotelling's T^{2} limit, scaled by the eigenvalue of the given PC(s). This is now consistent in that, if you mark any sample which is outside the 95% limit on any PC, this sample will also be outside the 95% limit for Hotelling's T^{2}.
UPDATE: A bug introduced in 3.5.2 with the change of limits was fixed in 3.5.4. The limits calculated by the ftest were not based on the correct number of components. Users should upgrade to 3.5.4 to correct this problem.
To see how to manulally calculate limits for scores, see the related FAQ.
Still having problems? Check our documentation Wiki or try writing our helpdesk at helpdesk@eigenvector.com